diff --git a/client/src/components/GTabs.vue b/client/src/components/GTabs.vue
index 3e15896..2e2d840 100644
--- a/client/src/components/GTabs.vue
+++ b/client/src/components/GTabs.vue
@@ -1,325 +1,325 @@
-<!-- This is a Galahad specific version of tabs -->
-<template>
-    <div class="tabs">
-        <!-- header -->
-        <div class="header">
-
-            <div class="top">
-                <div v-if="$slots['tabs-start']" class="nav tabs-start">
-                    <slot :name="'tabs-start'"></slot>
-                </div>
-                <slot :name="'title'"></slot>
-                <div v-if="$slots['tabs-end']" class="nav tabs-end">
-                    <slot :name="'tabs-end'"></slot>
-                </div>
-            </div>
-
-            <div class="bottom">
-                <div v-for="tab in _tabs" :key="tab.id">
-                    <a v-if="!tab.disabled && !tab.stub" disabled="disabled" :href='urlForTab(tab.id)'
-                        :class="'textcolor ' + navLinkClass(tab.id)" @click.prevent="navigateTo(tab.id)">
-                        <slot :name="`${tab.id}-title`" :isActive="currentTab === tab.id">{{ tab.title || tab.id }}
-                        </slot>
-                    </a>
-                    <span :class="`nav-link ${tab.disabled ? 'disabled' : ''}`" v-else>
-                        <slot :name="`${tab.id}-title`">{{ tab.title || tab.id }}</slot>
-                    </span>
-                </div>
-            </div>
-
-        </div>
-
-        <!-- content -->
-        <RouterView class="content" @navigate="x => { $router.push(x); induceCurrentTab() }" v-slot="{ Component }">
-            <transition name="fade" mode="out-in">
-                <component :is="Component" />
-            </transition>
-        </RouterView>
-
-    </div>
-</template>
-
-<script lang='ts'>
-import { defineComponent, PropType } from 'vue'
-
-// Stub means it does display as title, but it is not interactive, so you can implement custom functionality
-export type Tab = { id: string, title: string, disabled?: boolean, stub?: boolean }
-
-export default defineComponent({
-    name: "GTabs",
-    props: {
-        basePath: { type: String },
-        replace: { type: Boolean, default: false },
-        tabs: {
-            type: Array as PropType<Tab[]>,
-            // Ids to be used by the tabs. Pick anything that can be used as an attribute
-            default() {
-                return [{ id: 'id', title: 'You forgot the titles', disabled: true, stub: false }] as Tab[]
-            }
-        }
-    },
-    data() {
-        return {
-            currentTab: null // this.tabs[0].id
-        }
-    },
-    mounted() {
-        if (this.$route.path.split("/").length > this.basePath.split("/").length) {
-            // respect the url
-            this.induceCurrentTab()
-            return
-        }
-        const state = localStorage.getItem('galahad:' + this.basePath)
-        if (state !== null && state !== undefined) {
-            // load state from local storage
-            this.navigateTo(state, true)
-        } else {
-            // default
-            this.navigateTo(this._tabs[0].id, true)
-        }
-    },
-    computed: {
-        _tabs(): Tab[] {
-            // To allow for more flexible input, we calculate the actual tabs here
-            function isObject(objValue: unknown) {
-                return objValue && typeof objValue === 'object' && objValue.constructor === Object;
-            }
-            // Just assume it is either a correct object, or a string.
-            return this.tabs.map((x: Tab | string) => isObject(x) ? x : { id: x, title: x }) as Tab[]
-        },
-
-    },
-    methods: {
-        induceCurrentTab() {
-            const split = this.$route.path.split('/').reverse()
-            let induction = null // this._tabs[0].id // default
-            split.forEach(x => {
-                this._tabs.forEach(y => {
-                    if (x === y.id) induction = y.id
-                })
-            })
-            if (induction !== null)
-                this.setCurrentTab(induction) // TODO do this properly, this is bound to fail
-        },
-        navigateTo(tabId: string, replace = false) {
-            const path = this.basePath + '/' + tabId
-            if (!this.$route.path.startsWith(path)) {
-                if (replace || this.replace) {
-                    this.$router.replace({ path: path, query: this.$route.query })
-                } else {
-                    this.$router.push({ path: path, query: this.$route.query })
-                }
-            }
-            this.setCurrentTab(tabId)
-        },
-        navLinkClass(tabId: string) {
-            return "nav-link" + (this.currentTab == tabId ? " active" : "")
-        },
-        setCurrentTab(tabId: string) {
-            // Since the route is not reactive, we have to update the value like this
-            this.currentTab = tabId
-            if (tabId !== null && tabId !== undefined) localStorage.setItem('galahad:' + this.basePath, tabId)
-        },
-        urlForTab(tabId: string) {
-            const qs = Object.entries(this.$route.query).map(([k, v]) => `${k}=${encodeURIComponent(typeof (v) === "object" ? JSON.stringify(v) : v)}`).join('&')
-            return '/galahad' + this.basePath + '/' + tabId + '?' + qs;
-        }
-    },
-    watch: {
-        '$route'() {
-            this.induceCurrentTab()
-        }
-    }
-});
-</script>
-
-<style lang="scss" scoped>
-.tabs {
-    min-height: 0;
-    flex: 1;
-}
-
-.header {
-    position: -webkit-sticky;
-    /* Safari */
-    position: sticky;
-    font-family: Schoolboek, Helvetica, sans-serif;
-}
-
-// Header top
-.tabs>.header> :deep(.top) {
-    background-color: white;
-    align-items: center;
-    display: flex;
-    flex-wrap: wrap;
-
-    // header top tabs start
-    .tabs-start {
-
-        img {
-            position: relative;
-            left: -6px;
-            height: 125px;
-            pointer-events: none;
-        }
-
-        a {
-            display: block;
-            height: 110px;
-            width: 232px;
-            position: relative;
-        }
-    }
-
-    // header top title
-    .title {
-        display: inline-flex;
-        flex-direction: column;
-        padding: 0px 20px;
-
-        a {
-            font-style: normal;
-            color: black;
-            text-decoration: none;
-            font-size: 16px;
-
-            &:last-child {
-                font-size: 50px;
-                line-height: 50px;
-            }
-        }
-    }
-
-    // header top tabs end
-    .tabs-end {
-        flex: 1;
-        display: flex;
-        justify-content: flex-end;
-
-        a {
-            margin-right: 30px;
-            font-size: 16px;
-            color: black;
-            text-decoration: none;
-            font-style: normal;
-
-            &:hover {
-                text-decoration: underline;
-                cursor: pointer;
-            }
-        }
-    }
-}
-
-// Header bottom for all tabs
-.tabs .header .bottom {
-    display: inline-flex;
-    flex-wrap: wrap;
-    line-height: 50px;
-    background-color: var(--int-theme);
-
-    >div {
-        flex: 1 1 auto;
-    }
-
-    .nav-link {
-        font-style: normal;
-        text-align: center;
-        display: block;
-        min-width: 80px;
-        padding: 0 20px;
-        line-height: inherit;
-        text-decoration: none;
-        color: black;
-        user-select: none;
-        transition: background-color 0.1s linear;
-
-        &:hover:not(.disabled):not(.active) {
-            cursor: pointer;
-            background-color: var(--int-theme-hover);
-        }
-
-        &.disabled {
-            opacity: 0.5;
-            background-color: var(--int-very-light-grey);
-            // Same color as default css button:disabled
-            color: rgb(109, 109, 109);
-            cursor: not-allowed;
-        }
-
-        &.active {
-            background-color: var(--int-theme-outline);
-        }
-
-        &:active:not(.disabled):not(.active) {
-            background-color: var(--int-theme-active);
-        }
-    }
-}
-
-// tabs level 1
-.tabs.level-1 {
-    display: flex;
-    flex-direction: column;
-    height: 100%;
-
-    >.header {
-        background-color: var(--int-theme);
-        z-index: 100;
-        top: 0;
-        box-shadow: 0px 0px 10px black;
-
-        .top {
-            height: 110px;
-        }
-
-        .bottom {
-            line-height: 70px;
-            font-size: 18px;
-        }
-    }
-
-    >.content {
-        background-color: var(--int-very-light-grey);
-    }
-}
-
-/* It's nice to read with a bit more space at the bottom*/
-:deep(.tabs.level-2)>.content {
-    padding-bottom: 2em !important;
-}
-
-.tabs>.content {
-    padding: 10px;
-    min-height: 0;
-    display: flex;
-    flex-direction: column;
-    flex: 1;
-}
-
-.tabs.level-2,
-.tabs.level-3 {
-    box-sizing: border-box;
-
-    .content {
-        background-color: white;
-        border: var(--int-light-grey) 1px solid;
-    }
-
-    .bottom {
-        border-top: var(--int-light-grey) 1px solid;
-        border-left: var(--int-light-grey) 1px solid;
-        border-right: var(--int-light-grey) 1px solid;
-    }
-
-}
-
-.tabs.level-2 {
-    display: flex;
-    flex-direction: column;
-
-    .content {
-        overflow-y: auto;
-    }
-}
+<!-- This is a Galahad specific version of tabs -->
+<template>
+    <div class="tabs">
+        <!-- header -->
+        <div class="header">
+
+            <div class="top">
+                <div v-if="$slots['tabs-start']" class="nav tabs-start">
+                    <slot :name="'tabs-start'"></slot>
+                </div>
+                <slot :name="'title'"></slot>
+                <div v-if="$slots['tabs-end']" class="nav tabs-end">
+                    <slot :name="'tabs-end'"></slot>
+                </div>
+            </div>
+
+            <div class="bottom">
+                <div v-for="tab in _tabs" :key="tab.id">
+                    <a v-if="!tab.disabled && !tab.stub" disabled="disabled" :href='urlForTab(tab.id)'
+                        :class="'textcolor ' + navLinkClass(tab.id)" @click.prevent="navigateTo(tab.id)">
+                        <slot :name="`${tab.id}-title`" :isActive="currentTab === tab.id">{{ tab.title || tab.id }}
+                        </slot>
+                    </a>
+                    <span :class="`nav-link ${tab.disabled ? 'disabled' : ''}`" v-else>
+                        <slot :name="`${tab.id}-title`">{{ tab.title || tab.id }}</slot>
+                    </span>
+                </div>
+            </div>
+
+        </div>
+
+        <!-- content -->
+        <RouterView class="content" @navigate="x => { $router.push(x); induceCurrentTab() }" v-slot="{ Component }">
+            <transition name="fade" mode="out-in">
+                <component :is="Component" />
+            </transition>
+        </RouterView>
+
+    </div>
+</template>
+
+<script lang='ts'>
+import { defineComponent, PropType } from 'vue'
+
+// Stub means it does display as title, but it is not interactive, so you can implement custom functionality
+export type Tab = { id: string, title: string, disabled?: boolean, stub?: boolean }
+
+export default defineComponent({
+    name: "GTabs",
+    props: {
+        basePath: { type: String },
+        replace: { type: Boolean, default: false },
+        tabs: {
+            type: Array as PropType<Tab[]>,
+            // Ids to be used by the tabs. Pick anything that can be used as an attribute
+            default() {
+                return [{ id: 'id', title: 'You forgot the titles', disabled: true, stub: false }] as Tab[]
+            }
+        }
+    },
+    data() {
+        return {
+            currentTab: null // this.tabs[0].id
+        }
+    },
+    mounted() {
+        if (this.$route.path.split("/").length > this.basePath.split("/").length) {
+            // respect the url
+            this.induceCurrentTab()
+            return
+        }
+        const state = localStorage.getItem('galahad:' + this.basePath)
+        if (state !== null && state !== undefined) {
+            // load state from local storage
+            this.navigateTo(state, true)
+        } else {
+            // default
+            this.navigateTo(this._tabs[0].id, true)
+        }
+    },
+    computed: {
+        _tabs(): Tab[] {
+            // To allow for more flexible input, we calculate the actual tabs here
+            function isObject(objValue: unknown) {
+                return objValue && typeof objValue === 'object' && objValue.constructor === Object;
+            }
+            // Just assume it is either a correct object, or a string.
+            return this.tabs.map((x: Tab | string) => isObject(x) ? x : { id: x, title: x }) as Tab[]
+        },
+
+    },
+    methods: {
+        induceCurrentTab() {
+            const split = this.$route.path.split('/').reverse()
+            let induction = null // this._tabs[0].id // default
+            split.forEach(x => {
+                this._tabs.forEach(y => {
+                    if (x === y.id) induction = y.id
+                })
+            })
+            if (induction !== null)
+                this.setCurrentTab(induction) // TODO do this properly, this is bound to fail
+        },
+        navigateTo(tabId: string, replace = false) {
+            const path = this.basePath + '/' + tabId
+            if (!this.$route.path.startsWith(path)) {
+                if (replace || this.replace) {
+                    this.$router.replace({ path: path, query: this.$route.query })
+                } else {
+                    this.$router.push({ path: path, query: this.$route.query })
+                }
+            }
+            this.setCurrentTab(tabId)
+        },
+        navLinkClass(tabId: string) {
+            return "nav-link" + (this.currentTab == tabId ? " active" : "")
+        },
+        setCurrentTab(tabId: string) {
+            // Since the route is not reactive, we have to update the value like this
+            this.currentTab = tabId
+            if (tabId !== null && tabId !== undefined) localStorage.setItem('galahad:' + this.basePath, tabId)
+        },
+        urlForTab(tabId: string) {
+            const qs = Object.entries(this.$route.query).map(([k, v]) => `${k}=${encodeURIComponent(typeof (v) === "object" ? JSON.stringify(v) : v)}`).join('&')
+            return '/galahad' + this.basePath + '/' + tabId + '?' + qs;
+        }
+    },
+    watch: {
+        '$route'() {
+            this.induceCurrentTab()
+        }
+    }
+});
+</script>
+
+<style lang="scss" scoped>
+.tabs {
+    min-height: 0;
+    flex: 1;
+}
+
+.header {
+    position: -webkit-sticky;
+    /* Safari */
+    position: sticky;
+    font-family: Schoolboek, Helvetica, sans-serif;
+}
+
+// Header top
+.tabs>.header> :deep(.top) {
+    background-color: white;
+    align-items: center;
+    display: flex;
+    flex-wrap: wrap;
+
+    // header top tabs start
+    .tabs-start {
+
+        img {
+            position: relative;
+            left: -6px;
+            height: 125px;
+            pointer-events: none;
+        }
+
+        a {
+            display: block;
+            height: 110px;
+            width: 232px;
+            position: relative;
+        }
+    }
+
+    // header top title
+    .title {
+        display: inline-flex;
+        flex-direction: column;
+        padding: 0px 20px;
+
+        a {
+            font-style: normal;
+            color: black;
+            text-decoration: none;
+            font-size: 16px;
+
+            &:last-child {
+                font-size: 50px;
+                line-height: 50px;
+            }
+        }
+    }
+
+    // header top tabs end
+    .tabs-end {
+        flex: 1;
+        display: flex;
+        justify-content: flex-end;
+
+        a {
+            margin-right: 30px;
+            font-size: 16px;
+            color: black;
+            text-decoration: none;
+            font-style: normal;
+
+            &:hover {
+                text-decoration: underline;
+                cursor: pointer;
+            }
+        }
+    }
+}
+
+// Header bottom for all tabs
+.tabs .header .bottom {
+    display: inline-flex;
+    flex-wrap: wrap;
+    line-height: 50px;
+    background-color: var(--int-theme);
+
+    >div {
+        flex: 1 1 auto;
+    }
+
+    .nav-link {
+        font-style: normal;
+        text-align: center;
+        display: block;
+        min-width: 80px;
+        padding: 0 20px;
+        line-height: inherit;
+        text-decoration: none;
+        color: black;
+        user-select: none;
+        transition: background-color 0.1s linear;
+
+        &:hover:not(.disabled):not(.active) {
+            cursor: pointer;
+            background-color: var(--int-theme-hover);
+        }
+
+        &.disabled {
+            opacity: 0.5;
+            background-color: var(--int-very-light-grey);
+            // Same color as default css button:disabled
+            color: rgb(109, 109, 109);
+            cursor: not-allowed;
+        }
+
+        &.active {
+            background-color: var(--int-theme-outline);
+        }
+
+        &:active:not(.disabled):not(.active) {
+            background-color: var(--int-theme-active);
+        }
+    }
+}
+
+// tabs level 1
+.tabs.level-1 {
+    display: flex;
+    flex-direction: column;
+    height: 100%;
+
+    >.header {
+        background-color: var(--int-theme);
+        z-index: 1;
+        top: 0;
+        box-shadow: 0px 0px 10px black;
+
+        .top {
+            height: 110px;
+        }
+
+        .bottom {
+            line-height: 70px;
+            font-size: 18px;
+        }
+    }
+
+    >.content {
+        background-color: var(--int-very-light-grey);
+    }
+}
+
+/* It's nice to read with a bit more space at the bottom*/
+:deep(.tabs.level-2)>.content {
+    padding-bottom: 2em !important;
+}
+
+.tabs>.content {
+    padding: 10px;
+    min-height: 0;
+    display: flex;
+    flex-direction: column;
+    flex: 1;
+}
+
+.tabs.level-2,
+.tabs.level-3 {
+    box-sizing: border-box;
+
+    .content {
+        background-color: white;
+        border: var(--int-light-grey) 1px solid;
+    }
+
+    .bottom {
+        border-top: var(--int-light-grey) 1px solid;
+        border-left: var(--int-light-grey) 1px solid;
+        border-right: var(--int-light-grey) 1px solid;
+    }
+
+}
+
+.tabs.level-2 {
+    display: flex;
+    flex-direction: column;
+
+    .content {
+        overflow-y: auto;
+    }
+}
 </style>
\ No newline at end of file
diff --git a/client/src/components/input/UploadDocuments.vue b/client/src/components/input/UploadDocuments.vue
index 5ff21ea..6c8d9a7 100644
--- a/client/src/components/input/UploadDocuments.vue
+++ b/client/src/components/input/UploadDocuments.vue
@@ -1,107 +1,151 @@
-<template>
-    <div>
-
-        <!-- Styled label for input -->
-        <label for="file-upload" class="custom-file-upload">
-            <svg class="svg-icon" aria-hidden="true" role="img" focusable="false" xmlns="http://www.w3.org/2000/svg"
-                width="20" height="12" viewBox="0 0 20 12">
-                <polygon class="st0" points="10,4.2 2.2,12 0.1,9.9 10,0 19.9,9.9 17.8,12 "></polygon>
-            </svg>
-            Select file(s)
-        </label>
-        <!-- Actual input -->
-        <input type="file" ref="uploadInput" name="filefield" multiple id="file-upload" style="display: none;"
-            accept=".xml, .tsv, .txt, .zip, .conllu, .naf"
-            @change="e => filesToUpload = Object.values(e.target.files as FileList)" />
-
-
-        <!-- List of currently selected files. -->
-        <ul v-if="filesToUpload.length > 0">
-            <!-- First 5 file names are shown-->
-            <li v-for="file in filesToUpload.slice(0, 4)" :key="file.name">{{ file.name }}</li>
-            <!-- Overflow -->
-            <li v-if="filesToUpload.length > 4">
-                + {{ filesToUpload.length - 4 }} more {{ filesToUpload.length == 5 ? "file" : "files" }}
-            </li>
-        </ul>
-
-        <!-- Confirmation and clear buttons after a selection has been made -->
-        <template v-if="filesToUpload.length != 0">
-            <GButton green @click="documentsStore.uploadAll(); $refs.uploadInput.value = null">
-                Upload
-            </GButton>
-            <GButton plain @click="filesToUpload = []; $refs.uploadInput.value = null">
-                &#10006;&nbsp;clear
-            </GButton>
-        </template>
-
-        <!-- Error for illegal selection -->
-        <GInfo error v-if="illegalFiles.length > 0">
-            You have selected some filetype(s) that are not supported in GaLAHaD:
-            <ul>
-                <li v-for="file in illegalFiles" :key="file.name">
-                    {{ file.name }}
-                </li>
-            </ul>
-
-            Do you want to upload the text? You can:
-            <ol>
-                <li>copy-paste it to NotePad</li>
-                <li>save as .txt</li>
-                <li>upload it to GaLAHaD</li>
-            </ol>
-        </GInfo>
-
-        <!-- upload busy info -->
-        <GInfo v-show="uploadBusyCount > 0" spinner>
-            Upload will continue in the background, please don't close the browser. Currently uploading:
-            {{ uploadBusyCount }}
-        </GInfo>
-
-        <!-- Errors for files that could not be parsed by the server (e.g. broken xml tags)-->
-        <GInfo error v-show="uploadErrorCount > 0">
-            <div v-for="(value, key) in uploading" :key="key">
-                <span v-if="value.status == 'error'">{{ key }}: {{ value.message }}</span>
-            </div>
-        </GInfo>
-    </div>
-</template>
-
-<script setup lang='ts'>
-// Libraries & stores
-import stores from '@/stores'
-import { storeToRefs } from 'pinia'
-// Components
-import { GButton, GInfo } from '@/components'
-
-// Stores
-const documentsStore = stores.useDocuments()
-const { filesToUpload, illegalFiles, uploadBusyCount, uploadErrorCount, uploading } = storeToRefs(documentsStore)
-</script>
-
-<style scoped>
-.custom-file-upload {
-    border: 0px solid #ccc;
-    background-color: var(--int-theme);
-    display: inline-block;
-    padding: 10px 12px;
-    cursor: pointer;
-    margin-right: 10px;
-    font-style: normal;
-
-    &:hover {
-        background-color: var(--int-theme-hover);
-    }
-
-    &:active {
-        background-color: var(--int-theme-active);
-    }
-}
-
-ul,
-ol {
-    width: fit-content;
-    text-align: left;
-    margin: 1em auto;
-}
+<template>
+    <div>
+        <!-- Drag and drop from https://codepen.io/nekobog/pen/JjoZvBm -->
+        <div id="dropZone"></div>
+
+        <!-- Styled label for input -->
+        <label for="file-upload" class="custom-file-upload">
+            <svg class="svg-icon" aria-hidden="true" role="img" focusable="false" xmlns="http://www.w3.org/2000/svg"
+                width="20" height="12" viewBox="0 0 20 12">
+                <polygon class="st0" points="10,4.2 2.2,12 0.1,9.9 10,0 19.9,9.9 17.8,12 "></polygon>
+            </svg>
+            Select file(s) or drag & drop
+        </label>
+        <!-- Actual input -->
+        <input type="file" ref="uploadInput" name="filefield" multiple id="file-upload" style="display: none;"
+            accept=".xml, .tsv, .txt, .zip, .conllu, .naf"
+            @change="e => filesToUpload = Object.values(e.target.files as FileList)" />
+
+
+        <!-- List of currently selected files. -->
+        <ul v-if="filesToUpload.length > 0">
+            <!-- First 5 file names are shown-->
+            <li v-for="file in filesToUpload.slice(0, 4)" :key="file.name">{{ file.name }}</li>
+            <!-- Overflow -->
+            <li v-if="filesToUpload.length > 4">
+                + {{ filesToUpload.length - 4 }} more {{ filesToUpload.length == 5 ? "file" : "files" }}
+            </li>
+        </ul>
+
+        <!-- Confirmation and clear buttons after a selection has been made -->
+        <template v-if="filesToUpload.length != 0">
+            <GButton green @click="documentsStore.uploadAll(); $refs.uploadInput.value = null">
+                Upload
+            </GButton>
+            <GButton plain @click="filesToUpload = []; $refs.uploadInput.value = null">
+                &#10006;&nbsp;clear
+            </GButton>
+        </template>
+
+        <!-- Error for illegal selection -->
+        <GInfo error v-if="illegalFiles.length > 0">
+            You have selected some filetype(s) that are not supported in GaLAHaD:
+            <ul>
+                <li v-for="file in illegalFiles" :key="file.name">
+                    {{ file.name }}
+                </li>
+            </ul>
+
+            Do you want to upload the text? You can:
+            <ol>
+                <li>copy-paste it to NotePad</li>
+                <li>save as .txt</li>
+                <li>upload it to GaLAHaD</li>
+            </ol>
+        </GInfo>
+
+        <!-- upload busy info -->
+        <GInfo v-show="uploadBusyCount > 0" spinner>
+            Upload will continue in the background, please don't close the browser. Currently uploading:
+            {{ uploadBusyCount }}
+        </GInfo>
+
+        <!-- Errors for files that could not be parsed by the server (e.g. broken xml tags)-->
+        <GInfo error v-show="uploadErrorCount > 0">
+            <div v-for="(value, key) in uploading" :key="key">
+                <span v-if="value.status == 'error'">{{ key }}: {{ value.message }}</span>
+            </div>
+        </GInfo>
+    </div>
+</template>
+
+<script setup lang='ts'>
+// Libraries & stores
+import { onMounted, ref } from 'vue'
+import stores from '@/stores'
+import { storeToRefs } from 'pinia'
+// Components
+import { GButton, GInfo } from '@/components'
+
+// Stores
+const documentsStore = stores.useDocuments()
+const { filesToUpload, illegalFiles, uploadBusyCount, uploadErrorCount, uploading } = storeToRefs(documentsStore)
+
+// Fields
+var dropZone = ref(null as any as HTMLElement);
+
+// Methods
+function showDropZone(e: DragEvent) {
+    if (e.dataTransfer!.types.includes('Files')) {
+        dropZone.value.style.display = "block"
+    }
+}
+function hideDropZone() {
+    dropZone.value.style.display = "none";
+}
+function handleDrop(e: DragEvent) {
+    e.preventDefault();
+    hideDropZone();
+    filesToUpload.value = [...e.dataTransfer!.files]
+}
+
+// Watches & mounts
+onMounted(() => {
+    dropZone.value = document.getElementById('dropZone') as HTMLElement;
+    // Register drag events
+    window.addEventListener('dragenter', showDropZone)
+    dropZone.value.addEventListener('dragleave', hideDropZone)
+    dropZone.value.addEventListener('drop', handleDrop)
+    // Apparently, this is needed to prevent the browser from opening the file.
+    dropZone.value.addEventListener('dragover', e => e.preventDefault())
+})
+</script>
+
+<style scoped lang="scss">
+.custom-file-upload {
+    border: 0px solid #ccc;
+    background-color: var(--int-theme);
+    display: inline-block;
+    padding: 10px 12px;
+    cursor: pointer;
+    margin-right: 10px;
+    font-style: normal;
+
+    &:hover {
+        background-color: var(--int-theme-hover);
+    }
+
+    &:active {
+        background-color: var(--int-theme-active);
+    }
+}
+
+ul,
+ol {
+    width: fit-content;
+    text-align: left;
+    margin: 1em auto;
+}
+
+#dropZone {
+    display: none;
+    position: fixed;
+    top: 0;
+    left: 0;
+    width: 100%;
+    height: 100%;
+    background-color: var(--int-theme);
+    opacity: 0.4;
+    z-index: 3;
+}
 </style>
\ No newline at end of file
diff --git a/client/src/components/modals/GModal.vue b/client/src/components/modals/GModal.vue
index c8645ff..70a6465 100644
--- a/client/src/components/modals/GModal.vue
+++ b/client/src/components/modals/GModal.vue
@@ -1,108 +1,109 @@
-<template>
-  <div>
-
-    <transition name="fade" mode="out-in">
-      <!-- v-if instead of v-show such that elements inside a GModal can rely on onMounted()-->
-      <div v-if="show" class="bg" @click.self="$emit('hide')">
-        <GCard :showHelp="showHelp" :title="title" :class="`content ${small ? 'my-small' : ''}`" :noHelp="noHelp" :headless="headless">
-          <template #title>
-            <slot name="title"></slot>
-          </template>
-          <template #help>
-            <slot name="help"></slot>
-          </template>
-          <slot>
-            Somenone forgot to put the content
-          </slot>
-        </GCard>
-        <div class="buttons">
-          <GButton @click="$emit('hide')" red>Close</GButton>
-          <slot name="buttons"></slot>
-        </div>
-      </div>
-    </transition>
-
-  </div>
-</template>
-
-<script lang='ts'>
-import { defineComponent } from 'vue';
-
-export default defineComponent({
-  name: "GModal",
-  props: {
-    noHelp: { type: Boolean, default: false },
-    headless: { type: Boolean, default: false },
-    showHelp: { type: Boolean, default: true },
-    show: { type: Boolean },
-    small: { type: Boolean, default: false },
-    title: { type: String }
-  },
-  mounted(): void {
-    // smort
-    const alias = this.handleGlobalKeyDown
-    window.addEventListener('keyup', function (e) {
-      alias(e)
-    });
-  },
-  methods: {
-    handleGlobalKeyDown(e: KeyboardEvent): void {
-      if (e.key === "Escape") {
-        this.$emit('hide')
-      }
-    }
-  }
-});
-</script>
-
-<style scoped lang="scss">
-.bg {
-  left: 0;
-  top: 0;
-  background-color: var(--int-very-light-grey);
-  position: fixed;
-  height: 100%;
-  width: 100%;
-  z-index: 1000;
-  display: flex;
-  flex-direction: column;
-  box-sizing: border-box;
-  justify-content: center;
-  align-items: center;
-  padding: 2em;
-  padding-bottom: 1em; // Override for bottom button
-  gap: 1em;
-}
-
-.content {
-  overflow: auto;
-  border: 1px solid var(--int-light-grey);
-  box-sizing: border-box;
-  width: 100%;
-  padding: 2em;
-}
-
-.content.my-small {
-  // 'small' collides with a name from bootstrap(?). Hope te remove it in the future.
-  width: fit-content;
-}
-
-.content>* {
-  margin: 0px; // override default margin
-}
-
-.buttons {
-  display: flex;
-  justify-content: center;
-  box-sizing: border-box;
-  gap: 1em;
-}
-
-@media (max-width: 800px) or (max-height: 700px) {
-  .bg {
-    padding: 0;
-    padding-bottom: 0.5em; // Override for bottom button
-    gap: 0.5em;
-  }
-}
+<template>
+  <div>
+
+    <transition name="fade" mode="out-in">
+      <!-- v-if instead of v-show such that elements inside a GModal can rely on onMounted()-->
+      <div v-if="show" class="bg" @click.self="$emit('hide')">
+        <GCard :showHelp="showHelp" :title="title" :class="`content ${small ? 'my-small' : ''}`" :noHelp="noHelp"
+          :headless="headless">
+          <template #title>
+            <slot name="title"></slot>
+          </template>
+          <template #help>
+            <slot name="help"></slot>
+          </template>
+          <slot>
+            Somenone forgot to put the content
+          </slot>
+        </GCard>
+        <div class="buttons">
+          <GButton @click="$emit('hide')" red>Close</GButton>
+          <slot name="buttons"></slot>
+        </div>
+      </div>
+    </transition>
+
+  </div>
+</template>
+
+<script lang='ts'>
+import { defineComponent } from 'vue';
+
+export default defineComponent({
+  name: "GModal",
+  props: {
+    noHelp: { type: Boolean, default: false },
+    headless: { type: Boolean, default: false },
+    showHelp: { type: Boolean, default: true },
+    show: { type: Boolean },
+    small: { type: Boolean, default: false },
+    title: { type: String }
+  },
+  mounted(): void {
+    // smort
+    const alias = this.handleGlobalKeyDown
+    window.addEventListener('keyup', function (e) {
+      alias(e)
+    });
+  },
+  methods: {
+    handleGlobalKeyDown(e: KeyboardEvent): void {
+      if (e.key === "Escape") {
+        this.$emit('hide')
+      }
+    }
+  }
+});
+</script>
+
+<style scoped lang="scss">
+.bg {
+  left: 0;
+  top: 0;
+  background-color: var(--int-very-light-grey);
+  position: fixed;
+  height: 100%;
+  width: 100%;
+  z-index: 2;
+  display: flex;
+  flex-direction: column;
+  box-sizing: border-box;
+  justify-content: center;
+  align-items: center;
+  padding: 2em;
+  padding-bottom: 1em; // Override for bottom button
+  gap: 1em;
+}
+
+.content {
+  overflow: auto;
+  border: 1px solid var(--int-light-grey);
+  box-sizing: border-box;
+  width: 100%;
+  padding: 2em;
+}
+
+.content.my-small {
+  // 'small' collides with a name from bootstrap(?). Hope te remove it in the future.
+  width: fit-content;
+}
+
+.content>* {
+  margin: 0px; // override default margin
+}
+
+.buttons {
+  display: flex;
+  justify-content: center;
+  box-sizing: border-box;
+  gap: 1em;
+}
+
+@media (max-width: 800px) or (max-height: 700px) {
+  .bg {
+    padding: 0;
+    padding-bottom: 0.5em; // Override for bottom button
+    gap: 0.5em;
+  }
+}
 </style>
\ No newline at end of file
diff --git a/client/src/components/tables/GTable.vue b/client/src/components/tables/GTable.vue
index 2378169..3b70fb5 100644
--- a/client/src/components/tables/GTable.vue
+++ b/client/src/components/tables/GTable.vue
@@ -375,7 +375,7 @@ table.loading .loading-symbol {
 table .loading-symbol {
     transition: opacity 2s ease, visibility 2s ease;
     opacity: 0;
-    z-index: 3;
+    z-index: 1;
     position: absolute;
     top: 50%;
     left: 50%;
diff --git a/client/src/views/HomeView.vue b/client/src/views/HomeView.vue
index d400797..24af793 100644
--- a/client/src/views/HomeView.vue
+++ b/client/src/views/HomeView.vue
@@ -1,107 +1,111 @@
-<template>
-    <div>
-        <GCard noHelp>
-            <template #title>
-                Generating Linguistic Annotations for Historical Dutch
-            </template>
-
-            <hr>
-
-            <div class="flex">
-                <div class="left">
-                    <p>Historical texts are essential source material for both linguistic and digital humanities research.
-                        Adding linguistic annotation to historical text corpora helps to make the data more accessible.
-                        Users need not be concerned with historical spelling variation, and can query or analyse the data
-                        using higher-lever categories like part of speech and other grammatical properties.</p>
-                    <p>This application serves two purposes. One is to make annotation and tool evaluation easily accessible
-                        to researchers, the other to make it easy for developers to contribute their tools and models in the
-                        platform, and thus compare them to other tools with gold standard material included in the platform.
-                    </p>
-                    <p>GaLAHaD is designed to enable end users to choose the optimal path for their material. Apart from the
-                        basic task of uploading and annotating corpus material, GaLAHaD provides options to inspect and
-                        evaluate the result of the annotation process, in order to raise the awareness of typical errors and
-                        biases in the tools. The functionality of comparing annotation layers enables users to assess the
-                        accuracy of different tools on their data. It can be used both to evaluate a layer added by an
-                        automatic tagger with respect to a gold standard reference layer, or to compare layers added by
-                        different taggers. Disagreement between layers is not only represented by global statistics, but
-                        also illustrated by examples which are immediately visible in the tool. The annotated material can
-                        be automatically uploaded to the Autosearch corpus exploration environment and to the CoBaLT tool
-                        for manual correction of linguistic annotation.</p>
-                    <p>For tool developers, the docker-based application architecture ensures easy contribution of tools to
-                        the platform. The application and taggers are hosted by the INT and accessible with any
-                        CLARIN-account. There is also the option to self-host an instance using the publicly available
-                        docker images from the
-                        <ExternalLink href="https://hub.docker.com/u/instituutnederlandsetaal">INT docker hub
-                        </ExternalLink>
-                        or the open source code available on
-                        <ExternalLink href="https://www.github.com/INL/Galahad">GitHub</ExternalLink>.
-                    </p>
-                </div>
-
-                <figure class="right">
-                    <img src="@/assets/galahad-graal-medium.png" />
-                    <figcaption>
-                        <i>
-                            &ldquo;Ende seget datmen daer doe wan<br>
-                            Galaadde den goeden man,<br>
-                            Den besten riddre die wesen mochte,<br>
-                            Ende die ten inde brochte,<br>
-                            Nader historien tale,<br>
-                            Davonturen vanden grale&rdquo;<br>
-                        </i>
-                    </figcaption>
-                </figure>
-            </div>
-
-        </GCard>
-    </div>
-</template>
-
-<script setup lang='ts'>
-import { GCard, ExternalLink } from '@/components'
-</script>
-
-<style scoped="true" lang="scss">
-:deep(h3) {
-    font-size: 30px;
-    margin-bottom: 0;
-}
-
-.flex {
-    display: flex;
-    padding: 0 35px;
-    flex-wrap: wrap;
-    flex-direction: row;
-    justify-content: center;
-    gap: 35px;
-
-    .left {
-        flex: 2 1 500px;
-        max-width: 700px;
-    }
-
-    .right {
-        margin: 0;
-        margin-bottom: 3em;
-        flex: 1 2 300px;
-        max-width: 500px;
-        text-align: center;
-        font-family: 'Times New Roman', Times, serif;
-        font-size: 20px;
-
-        img {
-            object-fit: contain;
-            width: 100%;
-        }
-    }
-}
-
-
-.content {
-    background-color: white !important;
-}
-
-.g-card {
-    overflow-y: auto
-}
-</style>
+<template>
+    <div>
+        <GCard noHelp>
+            <template #title>
+                Generating Linguistic Annotations for Historical Dutch
+            </template>
+
+            <hr>
+
+            <div class="flex">
+                <div class="left">
+                    <p>
+                        Historical texts are essential source material for both linguistic and digital humanities
+                        research. Adding linguistic annotation (part of speech and modern Dutch lemma) to historical
+                        text helps to make the data more accessible. Users need not be concerned with historical
+                        spelling variation when querying and analysing the data.
+                    </p>
+                    <p>
+                        GaLAHaD serves two purposes. One is to make annotation and tool evaluation easily accessible to
+                        researchers, the other to make it easy for developers to contribute their tools and models to
+                        the platform, and thus compare them to other tools with gold standard material.
+                    </p>
+                    <p>
+                        <i>GaLAHaD</i> is designed to enable end users to choose the optimal path for their material.
+                        Apart from the basic task of uploading and annotating corpus material, <i>GaLAHaD</i> provides
+                        options to inspect and evaluate the result of the annotation process, in order to raise the
+                        awareness of typical errors and biases in the tools. The functionality of comparing annotation
+                        layers enables users to assess the accuracy of different tools on their data. It can be used
+                        both to evaluate a layer added by an automatic tagger with respect to a gold standard reference
+                        layer, or to compare layers added by different taggers. Disagreement between layers is not only
+                        represented by global statistics, but also illustrated by examples which are immediately visible
+                        in the tool. The annotated material can be uploaded to the Autosearch corpus exploration
+                        environment and to the <i>LAnCeLoT</i> tool for manual correction of linguistic annotation.
+                    </p>
+                    <p>
+                        For tool developers, the docker-based application architecture ensures easy contribution of
+                        tools to the platform. The application and taggers are hosted by the INT and accessible with any
+                        CLARIN-account. There is also the option to self-host an instance using the publicly available
+                        docker images from the
+                        <a href="https://hub.docker.com/u/instituutnederlandsetaal">INT Docker Hub</a> or the open
+                        source code available on <a href="https://github.com/INL/galahad">GitHub</a>.
+                    </p>
+                </div>
+
+                <figure class="right">
+                    <img src="@/assets/galahad-graal-medium.png" />
+                    <figcaption>
+                        <i>
+                            &ldquo;Ende seget datmen daer doe wan<br>
+                            Galaadde den goeden man,<br>
+                            Den besten riddre die wesen mochte,<br>
+                            Ende die ten inde brochte,<br>
+                            Nader historien tale,<br>
+                            Davonturen vanden grale&rdquo;<br>
+                        </i>
+                    </figcaption>
+                </figure>
+            </div>
+
+        </GCard>
+    </div>
+</template>
+
+<script setup lang='ts'>
+import { GCard, ExternalLink } from '@/components'
+</script>
+
+<style scoped="true" lang="scss">
+:deep(h3) {
+    font-size: 30px;
+    margin-bottom: 0;
+}
+
+.flex {
+    display: flex;
+    padding: 0 35px;
+    flex-wrap: wrap;
+    flex-direction: row;
+    justify-content: center;
+    gap: 35px;
+
+    .left {
+        flex: 2 1 500px;
+        max-width: 700px;
+    }
+
+    .right {
+        margin: 0;
+        margin-bottom: 3em;
+        flex: 1 2 300px;
+        max-width: 500px;
+        text-align: center;
+        font-family: 'Times New Roman', Times, serif;
+        font-size: 20px;
+
+        img {
+            object-fit: contain;
+            width: 100%;
+        }
+    }
+}
+
+
+.content {
+    background-color: white !important;
+}
+
+.g-card {
+    overflow-y: auto
+}
+</style>
diff --git a/client/src/views/help/subviews/GeneralView.vue b/client/src/views/help/subviews/GeneralView.vue
index 0c566c9..f3aeeac 100644
--- a/client/src/views/help/subviews/GeneralView.vue
+++ b/client/src/views/help/subviews/GeneralView.vue
@@ -132,8 +132,9 @@
         <p>
           For tool developers, the docker-based application architecture ensures easy contribution of tools to the
           platform. The application and taggers are hosted by the INT and accessible with any CLARIN-account. There is
-          also the option to self-host an instance using the publicly available docker images from the INT docker hub or
-          the open source code available on GitHub.
+          also the option to self-host an instance using the publicly available docker images from the
+          <a href="https://hub.docker.com/u/instituutnederlandsetaal">INT Docker Hub</a> or the open source code
+          available on <a href="https://github.com/INL/galahad">GitHub</a>.
         </p>
 
         <!-- H2: Release 1.0.0 -->
diff --git a/codemeta-harvest.json b/codemeta-harvest.json
index f9b0f51..7da42b6 100644
--- a/codemeta-harvest.json
+++ b/codemeta-harvest.json
@@ -1,389 +1,388 @@
-{
-    "@context": "https://w3id.org/codemeta/3.0",
-    "@type": "SoftwareSourceCode",
-    "dateCreated": "2024-05-31",
-    "datePublished": "2024-05-31",
-    "applicationCategory": [
-        "https://w3id.org/nwo-research-fields#Linguistics",
-        "https://w3id.org/nwo-research-fields#ComputationalLinguisticsAndPhilology",
-        "https://w3id.org/nwo-research-fields#ArtificialIntelligenceExpertSystems",
-        "https://w3id.org/nwo-research-fields#SoftwareForHumanities",
-        "https://w3id.org/nwo-research-fields#TextualAndLinguisticCorpora",
-        "https://vocabs.dariah.eu/tadirah/enriching",
-        "https://vocabs.dariah.eu/tadirah/posTagging",
-        "https://vocabs.dariah.eu/tadirah/tagging",
-        "https://vocabs.dariah.eu/tadirah/annotating",
-        "https://vocabs.dariah.eu/tadirah/merging",
-        "https://vocabs.dariah.eu/tadirah/converting",
-        "https://vocabs.dariah.eu/tadirah/lemmatizing",
-        "https://vocabs.dariah.eu/tadirah/machineLearning",
-        "https://vocabs.dariah.eu/tadirah/comparing",
-        "https://vocabs.dariah.eu/tadirah/analyzing"
-    ],
-    "author": [
-        {
-            "@type": "Person",
-            "@id": "http://orcid.org/0009-0006-9941-9582",
-            "email": "vincent.prins@ivdnt.org",
-            "familyName": "Prins",
-            "givenName": "Vincent"
-        },
-        {
-            "@type": "Person",
-            "familyName": "Brouwer",
-            "givenName": "Tim"
-        }
-    ],
-    "contributor": [
-        "http://orcid.org/0009-0006-9941-9582",
-        {
-            "@type": "Person",
-            "familyName": "Brouwer",
-            "givenName": "Tim"
-        }
-    ],
-    "maintainer": "http://orcid.org/0009-0006-9941-9582",
-    "codeRepository": "git+https://github.com/INL/galahad.git",
-    "description": "GaLAHaD (Generating Linguistic Annotations for Historical Dutch) allows linguists to compare taggers, tag their own corpora, evaluate the results and export their tagged documents.",
-    "downloadUrl": "https://github.com/INL/galahad",
-    "license": "http://spdx.org/licenses/Apache-2.0",
-    "name": "GaLAHaD",
-    "identifier": "galahad",
-    "funding": [
-        {
-            "@type": "Grant",
-            "name": "CLARIAH-PLUS (NWO grant 184.034.023)",
-            "funder": {
-                "@type": "Organization",
-                "name": "NWO",
-                "url": "https://www.nwo.nl"
-            }
-        }
-    ],
-    "operatingSystem": "Linux",
-    "programmingLanguage": [
-        "Kotlin",
-        "Javascript",
-        "Typescript"
-    ],
-    "runtimePlatform": [
-        "JVM",
-        "Node"
-    ],
-    "codemeta:contIntegration": {
-        "id": "https://github.com/INL/galahad/actions"
-    },
-    "continuousIntegration": "https://github.com/INL/galahad/actions",
-    "developmentStatus": [
-        "https://www.repostatus.org/#active",
-        "https://w3id.org/research-technology-readiness-levels#Level6LatePrototype"
-    ],
-    "producer": {
-        "@type": "Organization",
-        "name": "Instituut voor de Nederlandse taal",
-        "@id": "https://www.ivdnt.org",
-        "url": "https://www.ivdnt.org"
-    },
-    "issueTracker": "https://github.com/INL/galahad/issues",
-    "readme": "https://github.com/INL/Galahad/blob/release/readme.md",
-    "releaseNotes": "https://github.com/INL/Galahad/releases",
-    "softwareHelp": [
-        {
-            "@id": "https://portal.clarin.ivdnt.org/galahad/help",
-            "@type": "WebSite",
-            "name": "GaLAHad Help",
-            "url": "https://portal.clarin.ivdnt.org/galahad/help"
-        }
-    ],
-    "targetProduct": [
-        {
-            "@type": "WebApplication",
-            "name": "GaLAHaD",
-            "provider": "https://www.ivdnt.org",
-            "url": "https://portal.clarin.ivdnt.org/galahad"
-        },
-        {
-            "@type": "WebAPI",
-            "name": "GaLAHaD API",
-            "provider": "https://www.ivdnt.org",
-            "endpointUrl": {
-                "@type": "EntryPoint",
-                "url": "https://portal.clarin.ivdnt.org/galahad/api",
-                "contentType": "application/json"
-            },
-            "endpointDescription": {
-                "@type": "CreativeWork",
-                "encodingFormat": "application/json",
-                "url": "https://portal.clarin.ivdnt.org/galahad/api/v3/api-docs"
-            },
-            "documentation": "https://portal.clarin.ivdnt.org/galahad/api/swagger-ui/index.html",
-            "version": "1.0.0",
-            "consumesData": [
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "text/plain"
-                },
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "application/folia+xml"
-                },
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "application/tei+xml"
-                },
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "text/tab-separated-values"
-                },
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "https://github.com/newsreader/NAF"
-                },
-                {
-                    "@type": "TextDigitalDocument",
-                    "encodingFormat": "https://universaldependencies.org/format.html"
-                }
-            ],
-            "producesData": [
-                {
-                    "@type": "CreativeWork",
-                    "encodingFormat": "application/zip"
-                }
-            ]
-        },
-        {
-            "@type": "ServerApplication",
-            "name": "GaLAHaD proxy"
-        },
-        {
-            "@type": "SoftwareImage",
-            "name": "GaLAHaD client Docker image"
-        },
-        {
-            "@type": "SoftwareImage",
-            "name": "GaLAHaD server Docker image"
-        },
-        {
-            "@type": "SoftwareImage",
-            "name": "GaLAHaD proxy Docker image"
-        }
-    ],
-    "softwareRequirements": [
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@rollup/plugin-yaml",
-            "name": "@rollup/plugin-yaml",
-            "version": "^4.0.1"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@types/jest",
-            "name": "@types/jest",
-            "version": "^28.1.6"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@types/js-yaml",
-            "name": "@types/js-yaml",
-            "version": "^4.0.5"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@types/uuid",
-            "name": "@types/uuid",
-            "version": "^8.3.4"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@typescript-eslint/eslint-plugin",
-            "name": "@typescript-eslint/eslint-plugin",
-            "version": "^5.30.7"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@typescript-eslint/parser",
-            "name": "@typescript-eslint/parser",
-            "version": "^5.30.7"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@vitejs/plugin-vue",
-            "name": "@vitejs/plugin-vue",
-            "version": "^5.0.4"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "@vue/eslint-config-typescript",
-            "name": "@vue/eslint-config-typescript",
-            "version": "^11.0.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "axios",
-            "name": "axios",
-            "version": "^1.6.2"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "buffer",
-            "name": "buffer",
-            "version": "^6.0.3"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "content-disposition",
-            "name": "content-disposition",
-            "version": "^0.5.4"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "eslint",
-            "name": "eslint",
-            "version": "^8.20.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "eslint-plugin-vue",
-            "name": "eslint-plugin-vue",
-            "version": "^9.2.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "js-yaml",
-            "name": "js-yaml",
-            "version": "^4.1.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "json-loader",
-            "name": "json-loader",
-            "version": "^0.5.7"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "mutationobserver-shim",
-            "name": "mutationobserver-shim",
-            "version": "^0.3.7"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "node-sass",
-            "name": "node-sass",
-            "version": "^9.0.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "pinia",
-            "name": "pinia",
-            "version": "^2.0.16"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "safe-buffer",
-            "name": "safe-buffer",
-            "version": "^5.2.1"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "sass",
-            "name": "sass",
-            "version": "^1.53.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "typescript",
-            "name": "typescript",
-            "version": "^4.7.4"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "uuid",
-            "name": "uuid",
-            "version": "^8.3.2"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "vite",
-            "name": "vite",
-            "version": "^5.1.7"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "vue",
-            "name": "vue",
-            "version": "^3.2.37"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "vue-router",
-            "name": "vue-router",
-            "version": "^4.1.2"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "vue-slider-component",
-            "name": "vue-slider-component",
-            "version": "^4.1.0-beta.1"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "com.beust.klaxon",
-            "name": "klaxon",
-            "version": "5.6"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.jetbrains.kotlin.kotlin-reflect",
-            "name": "kotlin-reflect",
-            "version": "1.9.22"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.jetbrains.kotlin.kotlin-stdlib",
-            "name": "kotlin-stdlib",
-            "version": "1.9.22"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.jetbrains.kotlinx.kotlinx-coroutines-core-jvm",
-            "name": "kotlinx-coroutines-core-jvm",
-            "version": "1.7.3"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.jetbrains.kotlinx.kotlinx-serialization-json-jvm",
-            "name": "kotlinx-serialization-json-jvm",
-            "version": "1.6.3"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.apache.logging.log4j.log4j-api-kotlin",
-            "name": "log4j-api-kotlin",
-            "version": "1.2.0"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.yaml.snakeyaml",
-            "name": "snakeyaml",
-            "version": "2.2"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.springframework.boot.spring-boot-devtools",
-            "name": "spring-boot-devtools",
-            "version": "3.2.3"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.springframework.boot.spring-boot-starter-web",
-            "name": "spring-boot-starter-web",
-            "version": "3.2.4"
-        },
-        {
-            "@type": "SoftwareApplication",
-            "identifier": "org.springdoc.springdoc-openapi-starter-webmvc-ui",
-            "name": "springdoc-openapi-starter-webmvc-ui",
-            "version": "2.5.0"
-        }
-    ]
+{
+    "@context": "https://w3id.org/codemeta/3.0",
+    "@type": "SoftwareSourceCode",
+    "dateCreated": "2024-05-31",
+    "datePublished": "2024-05-31",
+    "applicationCategory": [
+        "https://w3id.org/nwo-research-fields#Linguistics",
+        "https://w3id.org/nwo-research-fields#ComputationalLinguisticsAndPhilology",
+        "https://w3id.org/nwo-research-fields#ArtificialIntelligenceExpertSystems",
+        "https://w3id.org/nwo-research-fields#SoftwareForHumanities",
+        "https://w3id.org/nwo-research-fields#TextualAndLinguisticCorpora",
+        "https://vocabs.dariah.eu/tadirah/enriching",
+        "https://vocabs.dariah.eu/tadirah/posTagging",
+        "https://vocabs.dariah.eu/tadirah/tagging",
+        "https://vocabs.dariah.eu/tadirah/annotating",
+        "https://vocabs.dariah.eu/tadirah/merging",
+        "https://vocabs.dariah.eu/tadirah/converting",
+        "https://vocabs.dariah.eu/tadirah/lemmatizing",
+        "https://vocabs.dariah.eu/tadirah/machineLearning",
+        "https://vocabs.dariah.eu/tadirah/comparing",
+        "https://vocabs.dariah.eu/tadirah/analyzing"
+    ],
+    "author": [
+        {
+            "@type": "Person",
+            "@id": "http://orcid.org/0009-0006-9941-9582",
+            "email": "vincent.prins@ivdnt.org",
+            "familyName": "Prins",
+            "givenName": "Vincent"
+        },
+        {
+            "@type": "Person",
+            "familyName": "Brouwer",
+            "givenName": "Tim"
+        }
+    ],
+    "contributor": [
+        "http://orcid.org/0009-0006-9941-9582",
+        {
+            "@type": "Person",
+            "familyName": "Brouwer",
+            "givenName": "Tim"
+        }
+    ],
+    "maintainer": "http://orcid.org/0009-0006-9941-9582",
+    "codeRepository": "git+https://github.com/INL/galahad.git",
+    "description": "GaLAHaD (Generating Linguistic Annotations for Historical Dutch) allows linguists to compare taggers, tag their own corpora, evaluate the results and export their tagged documents.",
+    "downloadUrl": "https://github.com/INL/galahad",
+    "license": "http://spdx.org/licenses/Apache-2.0",
+    "name": "GaLAHaD",
+    "identifier": "galahad",
+    "funding": [
+        {
+            "@type": "Grant",
+            "name": "CLARIAH-PLUS (NWO grant 184.034.023)",
+            "funder": {
+                "@type": "Organization",
+                "name": "NWO",
+                "url": "https://www.nwo.nl"
+            }
+        }
+    ],
+    "operatingSystem": "Linux",
+    "programmingLanguage": [
+        "Kotlin",
+        "Javascript",
+        "Typescript"
+    ],
+    "runtimePlatform": [
+        "JVM",
+        "Node"
+    ],
+    "codemeta:contIntegration": {
+        "id": "https://github.com/INL/galahad/actions"
+    },
+    "continuousIntegration": "https://github.com/INL/galahad/actions",
+    "developmentStatus": [
+        "https://www.repostatus.org/#active",
+        "https://w3id.org/research-technology-readiness-levels#Level6LatePrototype"
+    ],
+    "producer": {
+        "@type": "Organization",
+        "name": "Instituut voor de Nederlandse taal",
+        "@id": "https://www.ivdnt.org",
+        "url": "https://www.ivdnt.org"
+    },
+    "issueTracker": "https://github.com/INL/galahad/issues",
+    "readme": "https://github.com/INL/Galahad/blob/release/readme.md",
+    "releaseNotes": "https://github.com/INL/Galahad/releases",
+    "softwareHelp": [
+        {
+            "@id": "https://portal.clarin.ivdnt.org/galahad/help",
+            "@type": "WebSite",
+            "name": "GaLAHaD Help",
+            "url": "https://portal.clarin.ivdnt.org/galahad/help"
+        }
+    ],
+    "targetProduct": [
+        {
+            "@type": "WebApplication",
+            "name": "GaLAHaD",
+            "provider": "https://www.ivdnt.org",
+            "url": "https://portal.clarin.ivdnt.org/galahad"
+        },
+        {
+            "@type": "WebAPI",
+            "name": "GaLAHaD API",
+            "provider": "https://www.ivdnt.org",
+            "endpointUrl": {
+                "@type": "EntryPoint",
+                "url": "https://portal.clarin.ivdnt.org/galahad/api",
+                "contentType": "application/json"
+            },
+            "endpointDescription": {
+                "@type": "CreativeWork",
+                "encodingFormat": "application/json",
+                "url": "https://portal.clarin.ivdnt.org/galahad/api/v3/api-docs"
+            },
+            "documentation": "https://portal.clarin.ivdnt.org/galahad/api/swagger-ui/index.html",
+            "consumesData": [
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "text/plain"
+                },
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "application/folia+xml"
+                },
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "application/tei+xml"
+                },
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "text/tab-separated-values"
+                },
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "https://github.com/newsreader/NAF"
+                },
+                {
+                    "@type": "TextDigitalDocument",
+                    "encodingFormat": "https://universaldependencies.org/format.html"
+                }
+            ],
+            "producesData": [
+                {
+                    "@type": "CreativeWork",
+                    "encodingFormat": "application/zip"
+                }
+            ]
+        },
+        {
+            "@type": "ServerApplication",
+            "name": "GaLAHaD proxy"
+        },
+        {
+            "@type": "SoftwareImage",
+            "name": "GaLAHaD client Docker image"
+        },
+        {
+            "@type": "SoftwareImage",
+            "name": "GaLAHaD server Docker image"
+        },
+        {
+            "@type": "SoftwareImage",
+            "name": "GaLAHaD proxy Docker image"
+        }
+    ],
+    "softwareRequirements": [
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@rollup/plugin-yaml",
+            "name": "@rollup/plugin-yaml",
+            "version": "^4.0.1"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@types/jest",
+            "name": "@types/jest",
+            "version": "^28.1.6"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@types/js-yaml",
+            "name": "@types/js-yaml",
+            "version": "^4.0.5"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@types/uuid",
+            "name": "@types/uuid",
+            "version": "^8.3.4"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@typescript-eslint/eslint-plugin",
+            "name": "@typescript-eslint/eslint-plugin",
+            "version": "^5.30.7"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@typescript-eslint/parser",
+            "name": "@typescript-eslint/parser",
+            "version": "^5.30.7"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@vitejs/plugin-vue",
+            "name": "@vitejs/plugin-vue",
+            "version": "^5.0.4"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "@vue/eslint-config-typescript",
+            "name": "@vue/eslint-config-typescript",
+            "version": "^11.0.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "axios",
+            "name": "axios",
+            "version": "^1.6.2"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "buffer",
+            "name": "buffer",
+            "version": "^6.0.3"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "content-disposition",
+            "name": "content-disposition",
+            "version": "^0.5.4"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "eslint",
+            "name": "eslint",
+            "version": "^8.20.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "eslint-plugin-vue",
+            "name": "eslint-plugin-vue",
+            "version": "^9.2.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "js-yaml",
+            "name": "js-yaml",
+            "version": "^4.1.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "json-loader",
+            "name": "json-loader",
+            "version": "^0.5.7"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "mutationobserver-shim",
+            "name": "mutationobserver-shim",
+            "version": "^0.3.7"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "node-sass",
+            "name": "node-sass",
+            "version": "^9.0.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "pinia",
+            "name": "pinia",
+            "version": "^2.0.16"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "safe-buffer",
+            "name": "safe-buffer",
+            "version": "^5.2.1"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "sass",
+            "name": "sass",
+            "version": "^1.53.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "typescript",
+            "name": "typescript",
+            "version": "^4.7.4"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "uuid",
+            "name": "uuid",
+            "version": "^8.3.2"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "vite",
+            "name": "vite",
+            "version": "^5.1.7"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "vue",
+            "name": "vue",
+            "version": "^3.2.37"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "vue-router",
+            "name": "vue-router",
+            "version": "^4.1.2"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "vue-slider-component",
+            "name": "vue-slider-component",
+            "version": "^4.1.0-beta.1"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "com.beust.klaxon",
+            "name": "klaxon",
+            "version": "5.6"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.jetbrains.kotlin.kotlin-reflect",
+            "name": "kotlin-reflect",
+            "version": "1.9.22"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.jetbrains.kotlin.kotlin-stdlib",
+            "name": "kotlin-stdlib",
+            "version": "1.9.22"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.jetbrains.kotlinx.kotlinx-coroutines-core-jvm",
+            "name": "kotlinx-coroutines-core-jvm",
+            "version": "1.7.3"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.jetbrains.kotlinx.kotlinx-serialization-json-jvm",
+            "name": "kotlinx-serialization-json-jvm",
+            "version": "1.6.3"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.apache.logging.log4j.log4j-api-kotlin",
+            "name": "log4j-api-kotlin",
+            "version": "1.2.0"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.yaml.snakeyaml",
+            "name": "snakeyaml",
+            "version": "2.2"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.springframework.boot.spring-boot-devtools",
+            "name": "spring-boot-devtools",
+            "version": "3.2.3"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.springframework.boot.spring-boot-starter-web",
+            "name": "spring-boot-starter-web",
+            "version": "3.2.4"
+        },
+        {
+            "@type": "SoftwareApplication",
+            "identifier": "org.springdoc.springdoc-openapi-starter-webmvc-ui",
+            "name": "springdoc-openapi-starter-webmvc-ui",
+            "version": "2.5.0"
+        }
+    ]
 }
\ No newline at end of file
diff --git a/docker-compose.yml b/docker-compose.yml
index a80002d..a06c0fe 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -1,73 +1,79 @@
-# Docker compose file for galahad client, server and proxy
-services:
-    client:
-        image: instituutnederlandsetaal/galahad-client:${APP_VERSION}
-        restart: unless-stopped
-        networks:
-            - proxy_network
-        deploy:
-            resources:
-                limits:
-                    cpus: '0.50'
-                    memory: 512M
-                reservations:
-                    cpus: '0.25'
-                    memory: 128M
-    server:
-        image: instituutnederlandsetaal/galahad-server:${APP_VERSION}
-        restart: unless-stopped
-        volumes:
-            - user-corpora-volume:/data/corpora/custom
-            - preset-corpora-volume:/data/corpora/presets
-            - formats-volume:/data/formats
-            - taggers-volume:/data/taggers
-            - tagsets-volume:/data/tagsets
-            - admins-volume:/data/admins
-        networks:
-            - proxy_network
-            - taggers
-        deploy:
-            resources:
-                limits:
-                    cpus: '2.00'
-                    memory: 8G # should be higher that the java heap size
-                reservations:
-                    cpus: '1.00'
-                    memory: 2G
-        environment: # Otherwise container doesn't restart on OutOfMemory
-            - "JAVA_OPTS=-XX:+ExitOnOutOfMemoryError"
-    # Simple nginx reverse proxy to combine front- and backend
-    proxy:
-        image: instituutnederlandsetaal/galahad-proxy:${APP_VERSION}
-        depends_on:
-            - client
-            - server
-        ports:
-            - "80:80"
-        restart: unless-stopped
-        networks:
-            - proxy_network
-        deploy:
-            resources:
-                limits:
-                    cpus: '0.50'
-                    memory: 256M
-                reservations:
-                    cpus: '0.25'
-                    memory: 128M
-
-networks:
-    front:
-        driver: bridge
-    proxy_network:
-        driver: bridge
-    taggers:
-        external: true
-        name: $taggers_network
-volumes:
-    user-corpora-volume:
-    preset-corpora-volume:
-    formats-volume:
-    taggers-volume:
-    tagsets-volume:
-    admins-volume:
+# Docker compose file for galahad client, server and proxy
+services:
+    client:
+        image: instituutnederlandsetaal/galahad-client:${APP_VERSION}
+        build:
+            context: client
+        restart: unless-stopped
+        networks:
+            - proxy_network
+        deploy:
+            resources:
+                limits:
+                    cpus: '0.50'
+                    memory: 512M
+                reservations:
+                    cpus: '0.25'
+                    memory: 128M
+    server:
+        image: instituutnederlandsetaal/galahad-server:${APP_VERSION}
+        build:
+            context: server
+        restart: unless-stopped
+        volumes:
+            - user-corpora-volume:/data/corpora/custom
+            - preset-corpora-volume:/data/corpora/presets
+            - formats-volume:/data/formats
+            - taggers-volume:/data/taggers
+            - tagsets-volume:/data/tagsets
+            - admins-volume:/data/admins
+        networks:
+            - proxy_network
+            - taggers
+        deploy:
+            resources:
+                limits:
+                    cpus: '2.00'
+                    memory: 8G # should be higher that the java heap size
+                reservations:
+                    cpus: '1.00'
+                    memory: 2G
+        environment: # Otherwise container doesn't restart on OutOfMemory
+            - "JAVA_OPTS=-XX:+ExitOnOutOfMemoryError"
+    # Simple nginx reverse proxy to combine front- and backend
+    proxy:
+        image: instituutnederlandsetaal/galahad-proxy:${APP_VERSION}
+        build:
+            context: proxy
+        depends_on:
+            - client
+            - server
+        ports:
+            - "80:80"
+        restart: unless-stopped
+        networks:
+            - proxy_network
+        deploy:
+            resources:
+                limits:
+                    cpus: '0.50'
+                    memory: 256M
+                reservations:
+                    cpus: '0.25'
+                    memory: 128M
+
+networks:
+    front:
+        driver: bridge
+    proxy_network:
+        driver: bridge
+    taggers:
+        external: true
+        name: $taggers_network
+volumes:
+    user-corpora-volume:
+    preset-corpora-volume:
+    formats-volume:
+    taggers-volume:
+    tagsets-volume:
+    admins-volume:
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/FileBackedValue.kt b/server/src/main/kotlin/org/ivdnt/galahad/FileBackedValue.kt
index 791b1e8..8dccae2 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/FileBackedValue.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/FileBackedValue.kt
@@ -4,14 +4,8 @@ import com.fasterxml.jackson.core.type.TypeReference
 import com.fasterxml.jackson.databind.ObjectMapper
 import kotlinx.coroutines.Dispatchers
 import kotlinx.coroutines.runBlocking
-import kotlinx.coroutines.withContext
 import org.apache.logging.log4j.kotlin.Logging
-import java.io.*
-import java.nio.ByteBuffer
-import java.nio.channels.FileChannel
-import java.nio.channels.FileLock
-import java.nio.channels.OverlappingFileLockException
-import java.nio.file.StandardOpenOption
+import java.io.File
 
 
 val mapper: ObjectMapper by lazy { ObjectMapper() }
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/app/GalahadApplication.kt b/server/src/main/kotlin/org/ivdnt/galahad/app/GalahadApplication.kt
index ed9d009..b00c1e3 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/app/GalahadApplication.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/app/GalahadApplication.kt
@@ -1,214 +1,212 @@
-package org.ivdnt.galahad.app
-
-import com.fasterxml.jackson.annotation.JsonProperty
-import jakarta.servlet.http.HttpServletRequest
-import org.apache.logging.log4j.kotlin.Logging
-import org.apache.tomcat.util.http.fileupload.FileUploadException
-import org.springframework.beans.factory.annotation.Autowired
-import org.springframework.boot.autoconfigure.SpringBootApplication
-import org.springframework.boot.context.properties.ConfigurationProperties
-import org.springframework.boot.runApplication
-import org.springframework.boot.web.servlet.error.ErrorController
-import org.springframework.context.annotation.Bean
-import org.springframework.context.annotation.ComponentScan
-import org.springframework.context.annotation.Configuration
-import org.springframework.http.HttpStatus
-import org.springframework.http.ResponseEntity
-import org.springframework.web.bind.annotation.CrossOrigin
-import org.springframework.web.bind.annotation.GetMapping
-import org.springframework.web.bind.annotation.RequestMapping
-import org.springframework.web.bind.annotation.RestController
-import org.springframework.web.filter.CommonsRequestLoggingFilter
-import org.springframework.web.servlet.config.annotation.ViewControllerRegistry
-import org.springframework.web.servlet.config.annotation.WebMvcConfigurer
-import java.io.File
-import java.io.IOException
-import java.math.BigDecimal
-import java.net.URI
-import java.util.concurrent.TimeUnit
-import java.util.regex.Matcher
-import java.util.regex.Pattern
-
-// This is a possibly incomplete list of all the endpoints
-// For a complete overview better go to
-// SWAGGER_API_URL
-const val BASE_URL = "/"
-const val SWAGGER_API_URL = "/swagger-ui/index.html"
-
-const val TAGSETS_URL = "/tagsets"
-const val BENCHMARKS_URL = "/benchmarks"
-const val VERSION_URL = "/version"
-
-const val TAGGERS_URL = "/taggers"
-const val TAGGER_URL = "$TAGGERS_URL/{tagger}"
-const val TAGGER_HEALTH_URL = "$TAGGER_URL/health"
-
-const val ASSAYS_URL = "/assays"
-
-const val INTERNAL_JOBS_URL = "/internal/jobs"
-const val INTERNAL_JOBS_RESULT_URL = "$INTERNAL_JOBS_URL/result"
-const val INTERNAL_JOBS_ERROR_URL = "$INTERNAL_JOBS_URL/error"
-
-const val CORPORA_URL = "/corpora"
-const val PUBLIC_CORPORA_URL = "/public_corpora"
-const val DATASETS_CORPORA_URL = "/datasets_corpora"
-const val CORPUS_URL = "$CORPORA_URL/{corpus}"
-
-const val JOBS_URL = "$CORPUS_URL/jobs"
-const val JOB_URL = "$JOBS_URL/{job}"
-const val JOB_DOCUMENT_URL = "$JOB_URL/documents/{document}"
-
-const val EVALUATION_URL = "$JOB_URL/evaluation"
-const val ASSAY_URL = "$EVALUATION_URL/assay"
-const val DISTRIBUTION_URL = "$EVALUATION_URL/distribution"
-const val METRICS_URL = "$EVALUATION_URL/metrics"
-const val METRICS_CSV_URL = "$METRICS_URL/download"
-const val CONFUSION_URL = "$EVALUATION_URL/confusion"
-const val CONFUSION_CSV_URL = "$CONFUSION_URL/download"
-const val EVALUATION_CSV_URL = "$EVALUATION_URL/download"
-
-const val DOCUMENTS_URL = "$CORPUS_URL/documents"
-const val DOCUMENT_URL = "$DOCUMENTS_URL/{document}"
-const val DOCUMENT_RAW_FILE_URL = "$DOCUMENT_URL/raw" // returns the blob of the raw document
-
-var application_profile: String = System.getenv("spring.profiles.active") ?: "prod"
-fun String.runCommand(workingDir: File, timeout: Long = 60): String? {
-	try {
-		val parts = this.split("\\s".toRegex())
-		val proc = ProcessBuilder(*parts.toTypedArray())
-			.directory(workingDir)
-			.inheritIO()
-			.start()
-
-		proc.waitFor(timeout, TimeUnit.MINUTES)
-		return proc.inputStream.bufferedReader().readText()
-	} catch(e: IOException) {
-		e.printStackTrace()
-		return null
-	}
-}
-
-@Configuration
-@ConfigurationProperties(prefix = "")
-class Config {
-
-	lateinit var workDir: String
-
-	@Bean
-	fun getWorkingDirectory(): File {
-		return File( workDir )
-	}
-
-}
-
-@ComponentScan("org.ivdnt.galahad")
-@SpringBootApplication
-class GalahadApplication
-
-fun main(args: Array<String>) {
-	runApplication<GalahadApplication>(*args)
-}
-
-@RestController
-class ApplicationController : ErrorController, Logging {
-
-	@Autowired
-	private val request: HttpServletRequest? = null
-
-	@GetMapping( BASE_URL )
-	@CrossOrigin
-	fun getApplication(): ResponseEntity<Void> {
-		// Since we have nothing to show at this URL, we redirect to the API UI instead
-		logger.info( "Get root" )
-		return ResponseEntity.status(HttpStatus.FOUND).location(URI.create(request?.contextPath + SWAGGER_API_URL)).build()
-	}
-
-	@GetMapping( "/user" )
-	@CrossOrigin
-	fun getUser(): User {
-		logger.info( "Get user" )
-		return User.getUserFromRequestOrThrow( request )
-	}
-
-	@GetMapping( BENCHMARKS_URL )
-	@CrossOrigin
-	fun getBenchmarks(): ByteArray? {
-		logger.info( "Get benchmarks" )
-		val file = File( "benchmarks.yml" )
-		return if( file.exists() ) file.readBytes() else null
-	}
-
-	@GetMapping( VERSION_URL )
-	@CrossOrigin
-	fun getVersion(): ByteArray {
-		logger.info( "Get version" )
-		return this::class.java.classLoader.getResourceAsStream("version.yml")!!.readBytes()
-	}
-
-
-	data class ErrorResponse(
-		@JsonProperty val statusCode: HttpStatus,
-		@JsonProperty val message: String,
-	)
-
-	@RequestMapping("/error")
-//	@ResponseBody
-	@CrossOrigin
-	fun handleError(request: HttpServletRequest): ErrorResponse {
-		val statusCode = HttpStatus.valueOf( request.getAttribute("jakarta.servlet.error.status_code") as Int? ?: 500 )
-		val exception = request.getAttribute("jakarta.servlet.error.exception") as Exception?
-		if (exception?.cause is FileUploadException) {
-			// The error message is good as is. No need to wrap it
-			return ErrorResponse( statusCode, (exception.cause as FileUploadException).message ?: "exception inception" )
-		}
-		return ErrorResponse( statusCode, "${if (exception == null) "N/A" else exception.message}")
-	}
-
-}
-
-@Configuration
-@ConfigurationProperties(prefix = "spring.servlet.multipart")
-class MultipartConfig {
-
-	lateinit var maxFileSize: String
-	lateinit var maxRequestSize: String
-
-	val maxFilesSizeAsBytes: Long
-		get() {
-			return toBytes(maxFileSize)
-		}
-
-	companion object {
-		fun toBytes(filesize: String?): Long {
-			var returnValue: Long = -1
-			val patt: Pattern = Pattern.compile("([\\d.]+)([GMK]B)", Pattern.CASE_INSENSITIVE)
-			val matcher: Matcher = patt.matcher(filesize)
-			val powerMap: MutableMap<String, Int> = HashMap()
-			powerMap["GB"] = 3
-			powerMap["MB"] = 2
-			powerMap["KB"] = 1
-			if (matcher.find()) {
-				val number: String = matcher.group(1)
-				val pow = powerMap[matcher.group(2).uppercase()]!!
-				var bytes = BigDecimal(number)
-				bytes = bytes.multiply(BigDecimal.valueOf(1024).pow(pow))
-				returnValue = bytes.longValueExact()
-			}
-			return returnValue
-		}
-	}
-}
-
-@Configuration
-class RequestLoggingFilterConfig {
-	@Bean
-	fun logFilter(): CommonsRequestLoggingFilter {
-		val filter = CommonsRequestLoggingFilter()
-		filter.setIncludeQueryString(true)
-		filter.setIncludePayload(true)
-		filter.setMaxPayloadLength(10000)
-		filter.setIncludeHeaders(true)
-		filter.setAfterMessagePrefix("REQUEST DATA : ")
-		return filter
-	}
+package org.ivdnt.galahad.app
+
+import com.fasterxml.jackson.annotation.JsonProperty
+import jakarta.servlet.http.HttpServletRequest
+import org.apache.logging.log4j.kotlin.Logging
+import org.apache.tomcat.util.http.fileupload.FileUploadException
+import org.springframework.beans.factory.annotation.Autowired
+import org.springframework.boot.autoconfigure.SpringBootApplication
+import org.springframework.boot.context.properties.ConfigurationProperties
+import org.springframework.boot.runApplication
+import org.springframework.boot.web.servlet.error.ErrorController
+import org.springframework.context.annotation.Bean
+import org.springframework.context.annotation.ComponentScan
+import org.springframework.context.annotation.Configuration
+import org.springframework.http.HttpStatus
+import org.springframework.http.ResponseEntity
+import org.springframework.web.bind.annotation.CrossOrigin
+import org.springframework.web.bind.annotation.GetMapping
+import org.springframework.web.bind.annotation.RequestMapping
+import org.springframework.web.bind.annotation.RestController
+import org.springframework.web.filter.CommonsRequestLoggingFilter
+import java.io.File
+import java.io.IOException
+import java.math.BigDecimal
+import java.net.URI
+import java.util.concurrent.TimeUnit
+import java.util.regex.Matcher
+import java.util.regex.Pattern
+
+// This is a possibly incomplete list of all the endpoints
+// For a complete overview better go to
+// SWAGGER_API_URL
+const val BASE_URL = "/"
+const val SWAGGER_API_URL = "/swagger-ui/index.html"
+
+const val TAGSETS_URL = "/tagsets"
+const val BENCHMARKS_URL = "/benchmarks"
+const val VERSION_URL = "/version"
+
+const val TAGGERS_URL = "/taggers"
+const val TAGGER_URL = "$TAGGERS_URL/{tagger}"
+const val TAGGER_HEALTH_URL = "$TAGGER_URL/health"
+
+const val ASSAYS_URL = "/assays"
+
+const val INTERNAL_JOBS_URL = "/internal/jobs"
+const val INTERNAL_JOBS_RESULT_URL = "$INTERNAL_JOBS_URL/result"
+const val INTERNAL_JOBS_ERROR_URL = "$INTERNAL_JOBS_URL/error"
+
+const val CORPORA_URL = "/corpora"
+const val PUBLIC_CORPORA_URL = "/public_corpora"
+const val DATASETS_CORPORA_URL = "/datasets_corpora"
+const val CORPUS_URL = "$CORPORA_URL/{corpus}"
+
+const val JOBS_URL = "$CORPUS_URL/jobs"
+const val JOB_URL = "$JOBS_URL/{job}"
+const val JOB_DOCUMENT_URL = "$JOB_URL/documents/{document}"
+
+const val EVALUATION_URL = "$JOB_URL/evaluation"
+const val ASSAY_URL = "$EVALUATION_URL/assay"
+const val DISTRIBUTION_URL = "$EVALUATION_URL/distribution"
+const val METRICS_URL = "$EVALUATION_URL/metrics"
+const val METRICS_CSV_URL = "$METRICS_URL/download"
+const val CONFUSION_URL = "$EVALUATION_URL/confusion"
+const val CONFUSION_CSV_URL = "$CONFUSION_URL/download"
+const val EVALUATION_CSV_URL = "$EVALUATION_URL/download"
+
+const val DOCUMENTS_URL = "$CORPUS_URL/documents"
+const val DOCUMENT_URL = "$DOCUMENTS_URL/{document}"
+const val DOCUMENT_RAW_FILE_URL = "$DOCUMENT_URL/raw" // returns the blob of the raw document
+
+var application_profile: String = System.getenv("spring.profiles.active") ?: "prod"
+fun String.runCommand(workingDir: File, timeout: Long = 60): String? {
+	try {
+		val parts = this.split("\\s".toRegex())
+		val proc = ProcessBuilder(*parts.toTypedArray())
+			.directory(workingDir)
+			.inheritIO()
+			.start()
+
+		proc.waitFor(timeout, TimeUnit.MINUTES)
+		return proc.inputStream.bufferedReader().readText()
+	} catch(e: IOException) {
+		e.printStackTrace()
+		return null
+	}
+}
+
+@Configuration
+@ConfigurationProperties(prefix = "")
+class Config {
+
+	lateinit var workDir: String
+
+	@Bean
+	fun getWorkingDirectory(): File {
+		return File( workDir )
+	}
+
+}
+
+@ComponentScan("org.ivdnt.galahad")
+@SpringBootApplication
+class GalahadApplication
+
+fun main(args: Array<String>) {
+	runApplication<GalahadApplication>(*args)
+}
+
+@RestController
+class ApplicationController : ErrorController, Logging {
+
+	@Autowired
+	private val request: HttpServletRequest? = null
+
+	@GetMapping( BASE_URL )
+	@CrossOrigin
+	fun getApplication(): ResponseEntity<Void> {
+		// Since we have nothing to show at this URL, we redirect to the API UI instead
+		logger.info( "Get root" )
+		return ResponseEntity.status(HttpStatus.FOUND).location(URI.create(request?.contextPath + SWAGGER_API_URL)).build()
+	}
+
+	@GetMapping( "/user" )
+	@CrossOrigin
+	fun getUser(): User {
+		logger.info( "Get user" )
+		return User.getUserFromRequestOrThrow( request )
+	}
+
+	@GetMapping( BENCHMARKS_URL )
+	@CrossOrigin
+	fun getBenchmarks(): ByteArray? {
+		logger.info( "Get benchmarks" )
+		val file = File( "benchmarks.yml" )
+		return if( file.exists() ) file.readBytes() else null
+	}
+
+	@GetMapping( VERSION_URL )
+	@CrossOrigin
+	fun getVersion(): ByteArray {
+		logger.info( "Get version" )
+		return this::class.java.classLoader.getResourceAsStream("version.yml")!!.readBytes()
+	}
+
+
+	data class ErrorResponse(
+		@JsonProperty val statusCode: HttpStatus,
+		@JsonProperty val message: String,
+	)
+
+	@RequestMapping("/error")
+//	@ResponseBody
+	@CrossOrigin
+	fun handleError(request: HttpServletRequest): ErrorResponse {
+		val statusCode = HttpStatus.valueOf( request.getAttribute("jakarta.servlet.error.status_code") as Int? ?: 500 )
+		val exception = request.getAttribute("jakarta.servlet.error.exception") as Exception?
+		if (exception?.cause is FileUploadException) {
+			// The error message is good as is. No need to wrap it
+			return ErrorResponse( statusCode, (exception.cause as FileUploadException).message ?: "exception inception" )
+		}
+		return ErrorResponse( statusCode, "${if (exception == null) "N/A" else exception.message}")
+	}
+
+}
+
+@Configuration
+@ConfigurationProperties(prefix = "spring.servlet.multipart")
+class MultipartConfig {
+
+	lateinit var maxFileSize: String
+	lateinit var maxRequestSize: String
+
+	val maxFilesSizeAsBytes: Long
+		get() {
+			return toBytes(maxFileSize)
+		}
+
+	companion object {
+		fun toBytes(filesize: String?): Long {
+			var returnValue: Long = -1
+			val patt: Pattern = Pattern.compile("([\\d.]+)([GMK]B)", Pattern.CASE_INSENSITIVE)
+			val matcher: Matcher = patt.matcher(filesize)
+			val powerMap: MutableMap<String, Int> = HashMap()
+			powerMap["GB"] = 3
+			powerMap["MB"] = 2
+			powerMap["KB"] = 1
+			if (matcher.find()) {
+				val number: String = matcher.group(1)
+				val pow = powerMap[matcher.group(2).uppercase()]!!
+				var bytes = BigDecimal(number)
+				bytes = bytes.multiply(BigDecimal.valueOf(1024).pow(pow))
+				returnValue = bytes.longValueExact()
+			}
+			return returnValue
+		}
+	}
+}
+
+@Configuration
+class RequestLoggingFilterConfig {
+	@Bean
+	fun logFilter(): CommonsRequestLoggingFilter {
+		val filter = CommonsRequestLoggingFilter()
+		filter.setIncludeQueryString(true)
+		filter.setIncludePayload(true)
+		filter.setMaxPayloadLength(10000)
+		filter.setIncludeHeaders(true)
+		filter.setAfterMessagePrefix("REQUEST DATA : ")
+		return filter
+	}
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/app/report/Report.kt b/server/src/main/kotlin/org/ivdnt/galahad/app/report/Report.kt
index 08a64e3..6329c88 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/app/report/Report.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/app/report/Report.kt
@@ -2,8 +2,8 @@ package org.ivdnt.galahad.app.report
 
 import org.apache.logging.log4j.kotlin.Logging
 import org.apache.logging.log4j.kotlin.logger
-import org.ivdnt.galahad.data.layer.WordForm
 import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.data.layer.WordForm
 
 class Report : Logging {
 
@@ -18,8 +18,8 @@ class Report : Logging {
             // println("Spotted incompatible tokenization for \"${wf.literal}\" at offset ${wf.offset}")
             // Now we do nothing, but it is good to centrally register this
             logger().warn( "REPORT: Spotted incompatible tokenization for wordforms \n" +
-                    "    - $wf1 \n" +
-                    "    - $wf2"
+                    "    - ${wf1.literal} \n" +
+                    "    - ${wf2.literal}"
             )
         }
 
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/CorporaController.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/CorporaController.kt
index 10ee3a7..f7d6427 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/CorporaController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/CorporaController.kt
@@ -25,7 +25,7 @@ class CorporaController(
     @Autowired
     private val response: HttpServletResponse? = null
     private fun File.corpus(): Corpus {
-        return Corpus(this, User.getUserFromRequestOrThrow(request))
+        return Corpus(this)
     }
 
     private fun assertCorpusNameValidOrThrow(corpus: String) {
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/ExportController.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/ExportController.kt
index 8cc80b2..4bca5e2 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/ExportController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/ExportController.kt
@@ -1,159 +1,161 @@
-package org.ivdnt.galahad.data
-
-import jakarta.servlet.http.HttpServletRequest
-import jakarta.servlet.http.HttpServletResponse
-import org.apache.logging.log4j.kotlin.Logging
-import org.ivdnt.galahad.app.JOB_DOCUMENT_URL
-import org.ivdnt.galahad.app.JOB_URL
-import org.ivdnt.galahad.app.User
-import org.ivdnt.galahad.data.document.Document
-import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.port.CorpusTransformMetadata
-import org.ivdnt.galahad.port.DocumentTransformMetadata
-import org.ivdnt.galahad.port.InternalFile
-import org.ivdnt.galahad.util.setContentDisposition
-import org.springframework.beans.factory.annotation.Autowired
-import org.springframework.web.bind.annotation.*
-import java.io.File
-import java.util.*
-
-@RestController
-class ExportController(
-    val corpora: CorporaController,
-) : Logging {
-
-    @Autowired
-    private val request: HttpServletRequest? = null
-
-    @Autowired
-    private val response: HttpServletResponse? = null
-
-    private fun getCorpusTransformMetadata(corpusID: UUID, jobName: String): CorpusTransformMetadata {
-        // Exporting documents requires you to have write access.
-        val corpus = corpora.getWriteAccessOrThrow(corpusID, request)
-        val job = corpus.jobs.readOrThrow(jobName)
-        return CorpusTransformMetadata(
-            corpus = corpus, job = job, user = User.getUserFromRequestOrThrow(request)
-        )
-    }
-
-    private fun getDocumentTransformMetadata(
-        corpus: UUID,
-        job: String,
-        document: String,
-    ): DocumentTransformMetadata {
-        return getCorpusTransformMetadata(corpus, job).documentMetadata(document)
-    }
-
-    @GetMapping("$JOB_URL/export/convert")
-    @CrossOrigin
-    @ResponseBody
-    fun convertAndExportJob(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam("format") formatName: String,
-        @RequestParam("posHeadOnly") posHeadOnly: Boolean = false,
-    ) {
-        return exportCorpusJobInFormat(corpus, job, formatName, shouldMerge = false, posHeadOnly)
-    }
-
-    @GetMapping("$JOB_URL/export/merge")
-    @CrossOrigin
-    @ResponseBody
-    fun mergeAndExportJob(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam("format")
-        filterFormat: String, // Needed when raw formats are mixed, since BlackLab only accepts 1 type per corpus
-        @RequestParam("posHeadOnly")
-        posHeadOnly: Boolean,
-    ) {
-        return exportCorpusJobInFormat(corpus, job, filterFormat, shouldMerge = true, posHeadOnly)
-    }
-
-    fun exportCorpusJobInFormat(corpus: UUID, job: String, formatName: String, shouldMerge: Boolean, posHeadOnly: Boolean) {
-        val format = DocumentFormat.fromString(formatName)
-        val ctm = getCorpusTransformMetadata(corpus, job)
-        setZipResponseHeader(ctm)
-        ctm.corpus.getZipped(ctm, formatMapper = {
-            try {
-                // Document conversions.
-                val dtm = ctm.documentMetadata(it.name)
-                return@getZipped if (shouldMerge && mergeFormatMatches(it, format)) {
-                    logger.info("Merging ${it.name} of format ${it.format}")
-                    mergeAndExportDocument(dtm, posHeadOnly).file
-                } else {
-                    logger.info("Converting ${it.name} of format ${it.format} to $format")
-                    convertAndExportDocument(dtm, format, posHeadOnly)
-                }
-            } catch (e: Exception) {
-                throw Exception("Could not convert file ${it.name} to format ${format}. ${e.message}.")
-            }
-        }, filter = {
-            // Filter out untagged documents.
-                document ->
-            ctm.documentMetadata(document.name).layer != Layer.EMPTY
-        }, outputStream = response?.outputStream)
-    }
-
-    private fun mergeFormatMatches(
-        it: Document, format: DocumentFormat,
-    ): Boolean {
-        var otherFormat = it.format
-        // Overwrite the format for legacy formats that can in fact be merged.
-        if (otherFormat == DocumentFormat.TeiP5Legacy) {
-            otherFormat = DocumentFormat.TeiP5
-        }
-        return otherFormat == format
-    }
-
-    @GetMapping("$JOB_DOCUMENT_URL/export/convert")
-    @CrossOrigin
-    @ResponseBody
-    fun convertAndExportDocument(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @PathVariable document: String,
-        @RequestParam("format") formatName: String,
-        @RequestParam("posHeadOnly") posHeadOnly: Boolean,
-    ): ByteArray? {
-        val format = DocumentFormat.fromString(formatName)
-        val dtm = getDocumentTransformMetadata(corpus, job, document)
-        // TODO("set headers")
-        return convertAndExportDocument(dtm, format, posHeadOnly).readBytes()
-    }
-
-    fun convertAndExportDocument(dtm: DocumentTransformMetadata, format: DocumentFormat, posHeadOnly: Boolean): File {
-        if (posHeadOnly) {
-            dtm.convertLayerToPosHead()
-        }
-        return dtm.document.generateAs(format, dtm)
-    }
-
-    @GetMapping("$JOB_DOCUMENT_URL/export/merge")
-    @CrossOrigin
-    @ResponseBody
-    fun mergeAndExportDocument(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @PathVariable document: String,
-        @RequestParam("posHeadOnly") posHeadOnly: Boolean,
-    ): ByteArray? {
-        val dtm = getDocumentTransformMetadata(corpus, job, document)
-        // TODO("set headers")
-        return mergeAndExportDocument(dtm, posHeadOnly).file.readBytes()
-    }
-
-    fun mergeAndExportDocument(dtm: DocumentTransformMetadata, posHeadOnly: Boolean): InternalFile {
-        if (posHeadOnly) {
-            dtm.convertLayerToPosHead()
-        }
-        return dtm.document.merge(dtm)
-    }
-
-    private fun setZipResponseHeader(ctm: CorpusTransformMetadata) {
-        response!!.contentType = "application/zip"
-        response.setContentDisposition(ctm.corpus.metadata.expensiveGet().name + ".zip")
-    }
-}
+package org.ivdnt.galahad.data
+
+import jakarta.servlet.http.HttpServletRequest
+import jakarta.servlet.http.HttpServletResponse
+import org.apache.logging.log4j.kotlin.Logging
+import org.ivdnt.galahad.app.JOB_DOCUMENT_URL
+import org.ivdnt.galahad.app.JOB_URL
+import org.ivdnt.galahad.app.User
+import org.ivdnt.galahad.data.document.Document
+import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.port.CorpusTransformMetadata
+import org.ivdnt.galahad.port.DocumentTransformMetadata
+import org.ivdnt.galahad.port.InternalFile
+import org.ivdnt.galahad.util.setContentDisposition
+import org.springframework.beans.factory.annotation.Autowired
+import org.springframework.web.bind.annotation.*
+import java.io.File
+import java.util.*
+
+@RestController
+class ExportController(
+    val corpora: CorporaController,
+) : Logging {
+
+    @Autowired
+    private val request: HttpServletRequest? = null
+
+    @Autowired
+    private val response: HttpServletResponse? = null
+
+    private fun getCorpusTransformMetadata(corpusID: UUID, jobName: String, formatName: DocumentFormat): CorpusTransformMetadata {
+        // Exporting documents requires you to have write access.
+        val corpus = corpora.getWriteAccessOrThrow(corpusID, request)
+        val job = corpus.jobs.readOrThrow(jobName)
+        return CorpusTransformMetadata(
+            corpus = corpus, job = job, user = User.getUserFromRequestOrThrow(request), targetFormat = formatName
+        )
+    }
+
+    private fun getDocumentTransformMetadata(
+        corpus: UUID,
+        job: String,
+        document: String,
+        format: DocumentFormat,
+    ): DocumentTransformMetadata {
+        return getCorpusTransformMetadata(corpus, job, format).documentMetadata(document)
+    }
+
+    @GetMapping("$JOB_URL/export/convert")
+    @CrossOrigin
+    @ResponseBody
+    fun convertAndExportJob(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam("format") formatName: String,
+        @RequestParam("posHeadOnly") posHeadOnly: Boolean = false,
+    ) {
+        return exportCorpusJobInFormat(corpus, job, formatName, shouldMerge = false, posHeadOnly)
+    }
+
+    @GetMapping("$JOB_URL/export/merge")
+    @CrossOrigin
+    @ResponseBody
+    fun mergeAndExportJob(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam("format")
+        filterFormat: String, // Needed when raw formats are mixed, since BlackLab only accepts 1 type per corpus
+        @RequestParam("posHeadOnly")
+        posHeadOnly: Boolean,
+    ) {
+        return exportCorpusJobInFormat(corpus, job, filterFormat, shouldMerge = true, posHeadOnly)
+    }
+
+    fun exportCorpusJobInFormat(corpus: UUID, job: String, formatName: String, shouldMerge: Boolean, posHeadOnly: Boolean) {
+        val format = DocumentFormat.fromString(formatName)
+        val ctm = getCorpusTransformMetadata(corpus, job, format)
+        setZipResponseHeader(ctm)
+        ctm.corpus.getZipped(ctm, formatMapper = {
+            try {
+                // Document conversions.
+                val dtm = ctm.documentMetadata(it.name)
+                return@getZipped if (shouldMerge && mergeFormatMatches(it, format)) {
+                    logger.info("Merging ${it.name} of format ${it.format}")
+                    mergeAndExportDocument(dtm, posHeadOnly).file
+                } else {
+                    logger.info("Converting ${it.name} of format ${it.format} to $format")
+                    convertAndExportDocument(dtm, format, posHeadOnly)
+                }
+            } catch (e: Exception) {
+                throw Exception("Could not convert file ${it.name} to format ${format}. ${e.message}.")
+            }
+        }, filter = {
+            // Filter out untagged documents.
+                document ->
+            ctm.documentMetadata(document.name).layer != Layer.EMPTY
+        }, outputStream = response?.outputStream)
+    }
+
+    private fun mergeFormatMatches(
+        it: Document, format: DocumentFormat,
+    ): Boolean {
+        var otherFormat = it.format
+        // Overwrite the format for legacy formats that can in fact be merged.
+        if (otherFormat == DocumentFormat.TeiP5Legacy) {
+            otherFormat = DocumentFormat.TeiP5
+        }
+        return otherFormat == format
+    }
+
+    @GetMapping("$JOB_DOCUMENT_URL/export/convert")
+    @CrossOrigin
+    @ResponseBody
+    fun convertAndExportDocument(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @PathVariable document: String,
+        @RequestParam("format") formatName: String,
+        @RequestParam("posHeadOnly") posHeadOnly: Boolean,
+    ): ByteArray? {
+        val format = DocumentFormat.fromString(formatName)
+        val dtm = getDocumentTransformMetadata(corpus, job, document, format)
+        // TODO("set headers")
+        return convertAndExportDocument(dtm, format, posHeadOnly).readBytes()
+    }
+
+    fun convertAndExportDocument(dtm: DocumentTransformMetadata, format: DocumentFormat, posHeadOnly: Boolean): File {
+        if (posHeadOnly) {
+            dtm.convertLayerToPosHead()
+        }
+        return dtm.document.generateAs(format, dtm)
+    }
+
+    @GetMapping("$JOB_DOCUMENT_URL/export/merge")
+    @CrossOrigin
+    @ResponseBody
+    fun mergeAndExportDocument(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @PathVariable document: String,
+        @RequestParam("posHeadOnly") posHeadOnly: Boolean,
+    ): ByteArray? {
+        val doc = corpora.getWriteAccessOrThrow(corpus, request).documents.readOrThrow(document)
+        val dtm = getDocumentTransformMetadata(corpus, job, document, doc.format)
+        // TODO("set headers")
+        return mergeAndExportDocument(dtm, posHeadOnly).file.readBytes()
+    }
+
+    fun mergeAndExportDocument(dtm: DocumentTransformMetadata, posHeadOnly: Boolean): InternalFile {
+        if (posHeadOnly) {
+            dtm.convertLayerToPosHead()
+        }
+        return dtm.document.merge(dtm)
+    }
+
+    private fun setZipResponseHeader(ctm: CorpusTransformMetadata) {
+        response!!.contentType = "application/zip"
+        response.setContentDisposition(ctm.corpus.metadata.expensiveGet().name + ".zip")
+    }
+}
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/corpus/Corpus.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/corpus/Corpus.kt
index 382cd55..ba70eb3 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/corpus/Corpus.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/corpus/Corpus.kt
@@ -1,207 +1,206 @@
-package org.ivdnt.galahad.data.corpus
-
-import org.ivdnt.galahad.BaseFileSystemStore
-import org.ivdnt.galahad.FileBackedCache
-import org.ivdnt.galahad.FileBackedValue
-import org.ivdnt.galahad.app.ExpensiveGettable
-import org.ivdnt.galahad.app.User
-import org.ivdnt.galahad.app.executeAndLogTime
-import org.ivdnt.galahad.data.document.Document
-import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.document.Documents
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
-import org.ivdnt.galahad.jobs.Jobs
-import org.ivdnt.galahad.port.CmdiMetadata
-import org.ivdnt.galahad.port.CorpusTransformMetadata
-import org.ivdnt.galahad.taggers.Taggers
-import org.ivdnt.galahad.util.createZipFile
-import java.io.File
-import java.io.OutputStream
-import java.nio.file.Files
-import java.util.*
-import kotlin.io.path.createTempDirectory
-
-/**
- * A corpus is a collection of documents, metadata and jobs, saved to a folder. The folder contents are:
- *
- * - documents/: a folder containing all documents in the corpus. Represented by [Documents].
- * - jobs/: a folder containing all jobs that were active at some point in the corpus. The sourceLayer is one of them. Represented by [Jobs].
- * - metadata: a cache file storing [MutableCorpusMetadata] about the corpus.
- * - metadata.cache: a cache file storing [CorpusMetadata] about the corpus.
- *
- * A Corpus has an owner, who can add collaborators and viewers.
- * Collaborators have read and write access (and can add viewers and collaborators).
- * Viewers have read access.
- * Admins have access to all corpora with read and write access.
- */
-class Corpus(
-    workDirectory: File,
-    user: User,
-) : BaseFileSystemStore(workDirectory) {
-
-    val documents = Documents(workDirectory.resolve("documents"))
-
-    // Make sure this is initialized before accessing metadata
-    private val fileBackedMetadata: FileBackedValue<MutableCorpusMetadata>
-        get() = FileBackedValue(workDirectory.resolve("metadata"), MutableCorpusMetadata.initValue())
-
-    /**
-     * Convenient access to [MutableCorpusMetadata] without the need to get the expensive [CorpusMetadata]
-     * When uploading docs, for example, all we need to know is if the user has permission.
-     */
-    var mutableCorpusMetadata: MutableCorpusMetadata
-        get() = fileBackedMetadata.read<MutableCorpusMetadata>()
-        private set(value) = fileBackedMetadata.modify<MutableCorpusMetadata> { value }
-
-    private val metadataCache = object : FileBackedCache<CorpusMetadata>(
-        file = getMetadataFile(), initValue = CorpusMetadata()
-    ) {
-        override fun isValid(lastModified: Long): Boolean {
-            return lastModified >= fileBackedMetadata.lastModified
-        }
-
-        override fun set(): CorpusMetadata {
-            return CorpusMetadata(
-                // Mutable fields
-                owner = mutableCorpusMetadata.owner,
-                name = mutableCorpusMetadata.name,
-                eraTo = mutableCorpusMetadata.eraTo,
-                eraFrom = mutableCorpusMetadata.eraFrom,
-                tagset = mutableCorpusMetadata.tagset,
-                dataset = mutableCorpusMetadata.isDataset,
-                public = mutableCorpusMetadata.isDataset, // Note that we set isPublic the same as isDataset.
-                collaborators = mutableCorpusMetadata.collaborators ?: setOf(),
-                viewers = mutableCorpusMetadata.viewers ?: setOf(),
-                sourceName = mutableCorpusMetadata.sourceName,
-                sourceURL = mutableCorpusMetadata.sourceURL,
-                // Immutable/calculated fields
-                uuid = UUID.fromString(workDirectory.name),
-                activeJobs = jobs.readAll().filter { it.isActive }.size,
-                numDocs = documents.readAll().size,
-                sizeInBytes = workDirectory.walkTopDown().filter { it.isFile }.map { it.length() }.sum(), // expensive
-                lastModified = System.currentTimeMillis(),
-            )
-        }
-    }
-
-    private fun getMetadataFile() = workDirectory.resolve("metadata.cache")
-
-    /** Invalidate cache when new documents are uploaded or job activity changes */
-    fun invalidateCache() {
-        getMetadataFile().delete()
-    }
-
-    val metadata: ExpensiveGettable<CorpusMetadata> = object : ExpensiveGettable<CorpusMetadata> {
-        override fun expensiveGet() = metadataCache.get<CorpusMetadata>()
-    }
-
-    val sourceTagger: ExpensiveGettable<Taggers.Summary> = object : ExpensiveGettable<Taggers.Summary> {
-        override fun expensiveGet(): Taggers.Summary {
-            val metadata = metadata.expensiveGet()
-            return Taggers.Summary(
-                id = SOURCE_LAYER_NAME,
-                description = "uploaded annotations",
-                tagset = metadata.tagset,
-                eraFrom = metadata.eraFrom,
-                eraTo = metadata.eraTo,
-                produces = setOf("TODO"),
-            )
-        }
-    }
-
-    // Note: this is somewhat inefficient, since have to get the sourceTagger, even though we might not use it.
-    val jobs get() = Jobs(workDirectory.resolve("jobs"), this)
-
-    fun delete() {
-        workDirectory.deleteRecursively()
-    }
-
-    /**
-     * Overwrite the [CorpusMetadata] in [metadata] with [newMeta],
-     * except for the owner, which should be grabbed from the existing [metadata].
-     *
-     * If a user appears multiple times in the permission hierarchy, only the upper level remains.
-     */
-    fun updateMetadata(newMeta: MutableCorpusMetadata, user: User): ExpensiveGettable<CorpusMetadata> {
-        if (!mutableCorpusMetadata.isPublic && newMeta.isPublic) {
-            // Corpus is being set to public
-            if (!mutableCorpusMetadata.canMakePublic(user)) {
-                throw Exception("Unauthorized")
-            }
-        }
-        if (mutableCorpusMetadata.collaborators != newMeta.collaborators || mutableCorpusMetadata.viewers != newMeta.viewers) {
-            // Collaborators have changed
-            if (!mutableCorpusMetadata.canAddNewUsers(user) && mutableCorpusMetadata.owner != "") {
-                throw Exception("Unauthorized")
-            }
-        }
-        // If mutableCorpusMetadata.owner is "", we are working with the InitValue of FileBackedValue,
-        // so the updateMetadata call is initializing the corpus.
-        val owner = if (mutableCorpusMetadata.owner == "") user.id else mutableCorpusMetadata.owner
-        // Overwrite the owner with the original, so collaborators can't change it.
-        newMeta.owner = owner
-
-        // Trim textual intputs
-        newMeta.name = newMeta.name.trim()
-        newMeta.sourceName = newMeta.sourceName?.trim()
-        newMeta.tagset = newMeta.tagset?.trim()
-        newMeta.collaborators = newMeta.collaborators?.map { it.trim() }?.toSet()
-        newMeta.viewers = newMeta.viewers?.map { it.trim() }?.toSet()
-
-        // merge isPublic and isDataset
-        newMeta.isPublic = newMeta.isDataset
-
-        // Remove owner from list of collaborators & viewers
-        newMeta.collaborators = newMeta.collaborators?.filter { it != owner }?.toSet()
-        newMeta.viewers = newMeta.viewers?.filter { it != owner }?.toSet()
-        // Remove collaborators from list of viewers
-        if (newMeta.collaborators != null) newMeta.viewers = newMeta.viewers?.filter {
-            !newMeta.collaborators!!.contains(it)
-        }?.toSet()
-
-        mutableCorpusMetadata = newMeta
-        return metadata
-    }
-
-    fun removeAsViewer(user: User) {
-        fileBackedMetadata.modify<MutableCorpusMetadata> {
-            val viewers = it.viewers?.toMutableSet()
-            viewers?.removeIf { i -> i == user.id }
-            it.viewers = viewers
-            it
-        }
-    }
-
-    fun removeAsCollaborator(user: User) {
-        fileBackedMetadata.modify<MutableCorpusMetadata> {
-            val collaborators = it.collaborators?.toMutableSet()
-            collaborators?.removeIf { i -> i == user.id }
-            it.collaborators = collaborators
-            it
-        }
-    }
-
-    /**
-     * Maps all [Document] found in [Documents] to the desired [DocumentFormat] and zips them. [formatMapper] should perform the mapping.
-     */
-    fun getZipped(
-        ctm: CorpusTransformMetadata,
-        formatMapper: (Document) -> File,
-        filter: (Document) -> Boolean,
-        outputStream: OutputStream? = null,
-    ): File {
-        val name = metadata.expensiveGet().name
-        var zipFile: File? = null
-        val documents = documents.readAll().filter(filter)
-        executeAndLogTime("Generating $name zip") {
-            val convertedDocs = documents.asSequence().map(formatMapper)
-            val docsToCmdi = documents.asSequence().map { CmdiMetadata(ctm.documentMetadata(it.name)).file }
-            val cmdiZip = createZipFile(docsToCmdi, includeCMDI = true)
-            // rename the cmdiZip to "metadata"
-            val dest = File(createTempDirectory("metadata").toFile(), "metadata.zip")
-            Files.move(cmdiZip.toPath(), dest.toPath())
-            zipFile = createZipFile(convertedDocs + dest, outputStream)
-        }
-        return zipFile!!
-    }
+package org.ivdnt.galahad.data.corpus
+
+import org.ivdnt.galahad.BaseFileSystemStore
+import org.ivdnt.galahad.FileBackedCache
+import org.ivdnt.galahad.FileBackedValue
+import org.ivdnt.galahad.app.ExpensiveGettable
+import org.ivdnt.galahad.app.User
+import org.ivdnt.galahad.app.executeAndLogTime
+import org.ivdnt.galahad.data.document.Document
+import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.document.Documents
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
+import org.ivdnt.galahad.jobs.Jobs
+import org.ivdnt.galahad.port.CmdiMetadata
+import org.ivdnt.galahad.port.CorpusTransformMetadata
+import org.ivdnt.galahad.taggers.Taggers
+import org.ivdnt.galahad.util.createZipFile
+import java.io.File
+import java.io.OutputStream
+import java.nio.file.Files
+import java.util.*
+import kotlin.io.path.createTempDirectory
+
+/**
+ * A corpus is a collection of documents, metadata and jobs, saved to a folder. The folder contents are:
+ *
+ * - documents/: a folder containing all documents in the corpus. Represented by [Documents].
+ * - jobs/: a folder containing all jobs that were active at some point in the corpus. The sourceLayer is one of them. Represented by [Jobs].
+ * - metadata: a cache file storing [MutableCorpusMetadata] about the corpus.
+ * - metadata.cache: a cache file storing [CorpusMetadata] about the corpus.
+ *
+ * A Corpus has an owner, who can add collaborators and viewers.
+ * Collaborators have read and write access (and can add viewers and collaborators).
+ * Viewers have read access.
+ * Admins have access to all corpora with read and write access.
+ */
+class Corpus(
+    workDirectory: File,
+) : BaseFileSystemStore(workDirectory) {
+
+    val documents = Documents(workDirectory.resolve("documents"))
+
+    // Make sure this is initialized before accessing metadata
+    private val fileBackedMetadata: FileBackedValue<MutableCorpusMetadata>
+        get() = FileBackedValue(workDirectory.resolve("metadata"), MutableCorpusMetadata.initValue())
+
+    /**
+     * Convenient access to [MutableCorpusMetadata] without the need to get the expensive [CorpusMetadata]
+     * When uploading docs, for example, all we need to know is if the user has permission.
+     */
+    var mutableCorpusMetadata: MutableCorpusMetadata
+        get() = fileBackedMetadata.read<MutableCorpusMetadata>()
+        private set(value) = fileBackedMetadata.modify<MutableCorpusMetadata> { value }
+
+    private val metadataCache = object : FileBackedCache<CorpusMetadata>(
+        file = getMetadataFile(), initValue = CorpusMetadata()
+    ) {
+        override fun isValid(lastModified: Long): Boolean {
+            return lastModified >= fileBackedMetadata.lastModified
+        }
+
+        override fun set(): CorpusMetadata {
+            return CorpusMetadata(
+                // Mutable fields
+                owner = mutableCorpusMetadata.owner,
+                name = mutableCorpusMetadata.name,
+                eraTo = mutableCorpusMetadata.eraTo,
+                eraFrom = mutableCorpusMetadata.eraFrom,
+                tagset = mutableCorpusMetadata.tagset,
+                dataset = mutableCorpusMetadata.isDataset,
+                public = mutableCorpusMetadata.isDataset, // Note that we set isPublic the same as isDataset.
+                collaborators = mutableCorpusMetadata.collaborators ?: setOf(),
+                viewers = mutableCorpusMetadata.viewers ?: setOf(),
+                sourceName = mutableCorpusMetadata.sourceName,
+                sourceURL = mutableCorpusMetadata.sourceURL,
+                // Immutable/calculated fields
+                uuid = UUID.fromString(workDirectory.name),
+                activeJobs = jobs.readAll().filter { it.isActive }.size,
+                numDocs = documents.readAll().size,
+                sizeInBytes = workDirectory.walkTopDown().filter { it.isFile }.map { it.length() }.sum(), // expensive
+                lastModified = System.currentTimeMillis(),
+            )
+        }
+    }
+
+    private fun getMetadataFile() = workDirectory.resolve("metadata.cache")
+
+    /** Invalidate cache when new documents are uploaded or job activity changes */
+    fun invalidateCache() {
+        getMetadataFile().delete()
+    }
+
+    val metadata: ExpensiveGettable<CorpusMetadata> = object : ExpensiveGettable<CorpusMetadata> {
+        override fun expensiveGet() = metadataCache.get<CorpusMetadata>()
+    }
+
+    val sourceTagger: ExpensiveGettable<Taggers.Summary> = object : ExpensiveGettable<Taggers.Summary> {
+        override fun expensiveGet(): Taggers.Summary {
+            val metadata = metadata.expensiveGet()
+            return Taggers.Summary(
+                id = SOURCE_LAYER_NAME,
+                description = "uploaded annotations",
+                tagset = metadata.tagset,
+                eraFrom = metadata.eraFrom,
+                eraTo = metadata.eraTo,
+                produces = setOf("TODO"),
+            )
+        }
+    }
+
+    // Note: this is somewhat inefficient, since have to get the sourceTagger, even though we might not use it.
+    val jobs get() = Jobs(workDirectory.resolve("jobs"), this)
+
+    fun delete() {
+        workDirectory.deleteRecursively()
+    }
+
+    /**
+     * Overwrite the [CorpusMetadata] in [metadata] with [newMeta],
+     * except for the owner, which should be grabbed from the existing [metadata].
+     *
+     * If a user appears multiple times in the permission hierarchy, only the upper level remains.
+     */
+    fun updateMetadata(newMeta: MutableCorpusMetadata, user: User): ExpensiveGettable<CorpusMetadata> {
+        if (!mutableCorpusMetadata.isPublic && newMeta.isPublic) {
+            // Corpus is being set to public
+            if (!mutableCorpusMetadata.canMakePublic(user)) {
+                throw Exception("Unauthorized")
+            }
+        }
+        if (mutableCorpusMetadata.collaborators != newMeta.collaborators || mutableCorpusMetadata.viewers != newMeta.viewers) {
+            // Collaborators have changed
+            if (!mutableCorpusMetadata.canAddNewUsers(user) && mutableCorpusMetadata.owner != "") {
+                throw Exception("Unauthorized")
+            }
+        }
+        // If mutableCorpusMetadata.owner is "", we are working with the InitValue of FileBackedValue,
+        // so the updateMetadata call is initializing the corpus.
+        val owner = if (mutableCorpusMetadata.owner == "") user.id else mutableCorpusMetadata.owner
+        // Overwrite the owner with the original, so collaborators can't change it.
+        newMeta.owner = owner
+
+        // Trim textual intputs
+        newMeta.name = newMeta.name.trim()
+        newMeta.sourceName = newMeta.sourceName?.trim()
+        newMeta.tagset = newMeta.tagset?.trim()
+        newMeta.collaborators = newMeta.collaborators?.map { it.trim() }?.toSet()
+        newMeta.viewers = newMeta.viewers?.map { it.trim() }?.toSet()
+
+        // merge isPublic and isDataset
+        newMeta.isPublic = newMeta.isDataset
+
+        // Remove owner from list of collaborators & viewers
+        newMeta.collaborators = newMeta.collaborators?.filter { it != owner }?.toSet()
+        newMeta.viewers = newMeta.viewers?.filter { it != owner }?.toSet()
+        // Remove collaborators from list of viewers
+        if (newMeta.collaborators != null) newMeta.viewers = newMeta.viewers?.filter {
+            !newMeta.collaborators!!.contains(it)
+        }?.toSet()
+
+        mutableCorpusMetadata = newMeta
+        return metadata
+    }
+
+    fun removeAsViewer(user: User) {
+        fileBackedMetadata.modify<MutableCorpusMetadata> {
+            val viewers = it.viewers?.toMutableSet()
+            viewers?.removeIf { i -> i == user.id }
+            it.viewers = viewers
+            it
+        }
+    }
+
+    fun removeAsCollaborator(user: User) {
+        fileBackedMetadata.modify<MutableCorpusMetadata> {
+            val collaborators = it.collaborators?.toMutableSet()
+            collaborators?.removeIf { i -> i == user.id }
+            it.collaborators = collaborators
+            it
+        }
+    }
+
+    /**
+     * Maps all [Document] found in [Documents] to the desired [DocumentFormat] and zips them. [formatMapper] should perform the mapping.
+     */
+    fun getZipped(
+        ctm: CorpusTransformMetadata,
+        formatMapper: (Document) -> File,
+        filter: (Document) -> Boolean,
+        outputStream: OutputStream? = null,
+    ): File {
+        val name = metadata.expensiveGet().name
+        var zipFile: File? = null
+        val documents = documents.readAll().filter(filter)
+        executeAndLogTime("Generating $name zip") {
+            val convertedDocs = documents.asSequence().map(formatMapper)
+            val docsToCmdi = documents.asSequence().map { CmdiMetadata(ctm.documentMetadata(it.name)).file }
+            val cmdiZip = createZipFile(docsToCmdi, includeCMDI = true)
+            // rename the cmdiZip to "metadata"
+            val dest = File(createTempDirectory("metadata").toFile(), "metadata.zip")
+            Files.move(cmdiZip.toPath(), dest.toPath())
+            zipFile = createZipFile(convertedDocs + dest, outputStream)
+        }
+        return zipFile!!
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/document/Document.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/document/Document.kt
index 10b8112..b6ec005 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/document/Document.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/document/Document.kt
@@ -5,7 +5,10 @@ import org.ivdnt.galahad.FileBackedCache
 import org.ivdnt.galahad.FileBackedValue
 import org.ivdnt.galahad.app.ExpensiveGettable
 import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.port.*
+import org.ivdnt.galahad.port.DocumentTransformMetadata
+import org.ivdnt.galahad.port.InternalFile
+import org.ivdnt.galahad.port.PlainTextableFile
+import org.ivdnt.galahad.port.SourceLayerableFile
 import org.ivdnt.galahad.port.conllu.export.LayerToConlluConverter
 import org.ivdnt.galahad.port.folia.export.LayerToFoliaConverter
 import org.ivdnt.galahad.port.naf.export.LayerToNAFConverter
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/document/Documents.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/document/Documents.kt
index 680b66d..bccfe58 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/document/Documents.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/document/Documents.kt
@@ -3,7 +3,6 @@ package org.ivdnt.galahad.data.document
 import org.ivdnt.galahad.BaseFileSystemStore
 import org.ivdnt.galahad.app.CRUDSet
 import org.ivdnt.galahad.data.DocumentWriteType
-import org.ivdnt.galahad.data.DocumentsController
 import java.io.File
 
 /**
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/data/layer/Term.kt b/server/src/main/kotlin/org/ivdnt/galahad/data/layer/Term.kt
index 2aeddd8..66e63ad 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/data/layer/Term.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/data/layer/Term.kt
@@ -1,102 +1,112 @@
-package org.ivdnt.galahad.data.layer
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import com.fasterxml.jackson.annotation.JsonProperty
-
-/** Avoid empty strings in the CSV representation. */
-fun Term.toNonEmptyPair(): Pair<String, String> {
-    return (this.pos ?: Term.NO_POS) to (this.lemma ?: Term.NO_LEMMA)
-}
-
-/**
- * A term in a [Layer]. A term has a [lemma], a [pos] and refers to one or multiple [WordForm].
- * Referring to multiple [WordForm] is used to represent multi-word terms, although it is currently not used.
- * Lemma and pos can be null.
- */
-data class Term(
-    @JsonProperty("lemma") val lemma: String?,
-    @JsonProperty("pos") val pos: String?,
-    @JsonProperty("targets") val targets: MutableList<WordForm>,
-) {
-    /** Whether the lemma is not null. */
-    @get:JsonIgnore
-    val hasLemma: Boolean = lemma != null
-
-    /** Whether the pos is not null. */
-    @get:JsonIgnore
-    val hasPOS: Boolean = pos != null
-
-    @get:JsonIgnore
-    val posHeadGroupOrDefault
-        get() = posHeadGroup ?: NO_POS
-
-    @get:JsonIgnore
-    val lemmaOrDefault
-        get() = lemma ?: NO_LEMMA
-
-    /** Whether this term refers to multiple [WordForm]. */
-    @get:JsonIgnore
-    val isMultiTarget = targets.size > 1
-
-    /** The head of the first [pos]. E.g. "PD" for "PD(type=art)+NOU(num=sg)". */
-    @get:JsonIgnore
-    val posHead: String? = posToPosHead(pos)
-
-    @get:JsonIgnore
-    val isMultiPos: Boolean = pos?.contains("+") ?: false
-
-    /** The head of all [pos]. E.g. "PD+NOU" for "PD(type=art)+NOU(num=sg)". */
-    @get:JsonIgnore
-    val posHeadGroup: String? = run {
-            // Split on +
-            if (!isMultiPos) return@run posHead
-            val result: String? = pos?.split("+")?.mapNotNull { posToPosHead(it) }?.joinToString("+")
-            result
-        }
-
-
-
-    /** The features of [pos]. E.g. "num=sg" for "NOU(num=sg)". Does not support multi-pos. */
-    @get:JsonIgnore
-    val posFeatures: String?
-        get() {
-            if (pos == null) return null
-            val featureStart: Int = pos?.indexOf('(') ?: -1
-            val featureEnd: Int = pos?.indexOf(')') ?: -1
-            return if (featureStart != -1 && featureEnd != -1) {
-                return pos!!.slice(featureStart + 1 until featureEnd)
-            } else null
-        }
-
-    /** Offset of the first [WordForm] in [targets].*/
-    @get:JsonIgnore
-    val firstOffset get() = targets.minOfOrNull { it.offset } ?: -1
-
-    /** String constructed from all the [WordForm] in [targets]. */
-    @get:JsonIgnore
-    val literals: String
-        get() = targets.joinToString(" ") { it.literal }
-
-    companion object {
-        const val NO_POS = "NO_POS"
-        const val NO_LEMMA = "NO_LEMMA"
-        val EMPTY = Term(null, null, mutableListOf())
-        private fun posToPosHead(pos: String?): String? {
-            return if (pos == null) {
-                null
-            } else if (pos.contains('(')) {
-                // pos contains a non-letter non-digit character
-                val headEnd = pos.indexOf('(')
-                val head = pos.slice(0 until headEnd)
-                if (head.isEmpty()) {
-                    pos // pos is non-empty and starts with a non-letter character, e.g.: _
-                } else {
-                    head
-                }
-            } else {
-                // pos is 0 or more letters only
-                pos
-            }
-        }
-    }
+package org.ivdnt.galahad.data.layer
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import com.fasterxml.jackson.annotation.JsonProperty
+
+/** Avoid empty strings in the CSV representation. */
+fun Term.toNonEmptyPair(): Pair<String, String> {
+    return (this.pos ?: Term.NO_POS) to (this.lemma ?: Term.NO_LEMMA)
+}
+
+/**
+ * A term in a [Layer]. A term has a [lemma], a [pos] and refers to one or multiple [WordForm].
+ * Referring to multiple [WordForm] is used to represent multi-word terms, although it is currently not used.
+ * Lemma and pos can be null.
+ */
+data class Term(
+    @JsonProperty("lemma") val lemma: String?,
+    @JsonProperty("pos") val pos: String?,
+    @JsonProperty("targets") val targets: MutableList<WordForm>,
+) {
+    /** Whether the lemma is not null. */
+    @get:JsonIgnore
+    val hasLemma: Boolean = lemma != null
+
+    /** Whether the pos is not null. */
+    @get:JsonIgnore
+    val hasPOS: Boolean = pos != null
+
+    @get:JsonIgnore
+    val posHeadGroupOrDefault
+        get() = posHeadGroup ?: NO_POS
+
+    @get:JsonIgnore
+    val lemmaOrDefault
+        get() = lemma ?: NO_LEMMA
+
+    @get:JsonIgnore
+    val lemmaOrEmpty
+        get() = lemma ?: ""
+
+    @get:JsonIgnore
+    val posOrEmpty
+        get() = pos ?: ""
+
+    /** Whether this term refers to multiple [WordForm]. */
+    @get:JsonIgnore
+    val isMultiTarget = targets.size > 1
+
+    /** The head of the first [pos]. E.g. "PD" for "PD(type=art)+NOU(num=sg)". */
+    @get:JsonIgnore
+    val posHead: String? = posToPosHead(pos)
+
+    @get:JsonIgnore
+    val isMultiPos: Boolean = pos?.contains("+") ?: false
+
+    /** The head of all [pos]. E.g. "PD+NOU" for "PD(type=art)+NOU(num=sg)". */
+    @get:JsonIgnore
+    val posHeadGroup: String? = run {
+        // Split on +
+        if (!isMultiPos) return@run posHead
+        val result: String? = pos?.split("+")?.mapNotNull { posToPosHead(it) }?.joinToString("+")
+        result
+    }
+
+    @get:JsonIgnore
+    val posHeadGroupOrEmpty
+        get() = posHeadGroup ?: ""
+
+    /** The features of [pos]. E.g. "num=sg" for "NOU(num=sg)". Does not support multi-pos. */
+    @get:JsonIgnore
+    val posFeatures: String?
+        get() {
+            if (pos == null) return null
+            val featureStart: Int = pos.indexOf('(')
+            val featureEnd: Int = pos.indexOf(')')
+            return if (featureStart != -1 && featureEnd != -1) {
+                return pos.slice(featureStart + 1 until featureEnd)
+            } else null
+        }
+
+    /** Offset of the first [WordForm] in [targets].*/
+    @get:JsonIgnore
+    val firstOffset get() = targets.minOfOrNull { it.offset } ?: -1
+
+    /** String constructed from all the [WordForm] in [targets]. */
+    @get:JsonIgnore
+    val literals: String
+        get() = targets.joinToString(" ") { it.literal }
+
+    companion object {
+        const val NO_POS = "NO_POS"
+        const val NO_LEMMA = "NO_LEMMA"
+        val EMPTY = Term(null, null, mutableListOf())
+        private fun posToPosHead(pos: String?): String? {
+            return if (pos == null) {
+                null
+            } else if (pos.contains('(')) {
+                // pos contains a non-letter non-digit character
+                val headEnd = pos.indexOf('(')
+                val head = pos.slice(0 until headEnd)
+                if (head.isEmpty()) {
+                    pos // pos is non-empty and starts with a non-letter character, e.g.: _
+                } else {
+                    head
+                }
+            } else {
+                // pos is 0 or more letters only
+                pos
+            }
+        }
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/EvaluationController.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/EvaluationController.kt
index 1eb858f..eed3d90 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/EvaluationController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/EvaluationController.kt
@@ -1,256 +1,256 @@
-package org.ivdnt.galahad.evaluation
-
-import jakarta.servlet.http.HttpServletRequest
-import jakarta.servlet.http.HttpServletResponse
-import org.apache.logging.log4j.kotlin.Logging
-import org.ivdnt.galahad.app.*
-import org.ivdnt.galahad.data.CorporaController
-import org.ivdnt.galahad.data.corpus.CorpusMetadata
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
-import org.ivdnt.galahad.evaluation.comparison.ConfusionLayerFilter
-import org.ivdnt.galahad.evaluation.comparison.PosLemmaTermFilter
-import org.ivdnt.galahad.evaluation.comparison.MetricsLayerFilter
-import org.ivdnt.galahad.evaluation.confusion.Confusion
-import org.ivdnt.galahad.evaluation.confusion.CorpusConfusion
-import org.ivdnt.galahad.evaluation.distribution.CorpusDistribution
-import org.ivdnt.galahad.evaluation.metrics.CorpusMetrics
-import org.ivdnt.galahad.evaluation.metrics.MetricsSettings
-import org.ivdnt.galahad.evaluation.metrics.METRIC_TYPES
-import org.ivdnt.galahad.evaluation.metrics.PosByPosMetricsSettings
-import org.ivdnt.galahad.port.csv.CSVFile
-import org.ivdnt.galahad.util.createZipFile
-import org.ivdnt.galahad.util.setContentDisposition
-import org.springframework.beans.factory.annotation.Autowired
-import org.springframework.web.bind.annotation.*
-import java.io.File
-import java.time.LocalDateTime
-import java.time.format.DateTimeFormatter
-import java.util.*
-import kotlin.io.path.createTempDirectory
-
-const val DISTRIBUTION_MAX_SIZE = 1000
-
-@RestController
-class EvaluationController(
-    val corpora: CorporaController,
-) : Logging {
-
-    @Autowired
-    private val request: HttpServletRequest? = null
-
-    @Autowired
-    private val response: HttpServletResponse? = null
-
-    @GetMapping(DISTRIBUTION_URL)
-    @CrossOrigin
-    fun getDistribution(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-    ): CorpusDistribution {
-        logger.info("Get distribution for hypothesis layer $job in $corpus")
-        return CorpusDistribution(
-            corpora.getReadAccessOrThrow(corpus, request),
-            job
-        ).trim(DISTRIBUTION_MAX_SIZE) as CorpusDistribution
-    }
-
-    @GetMapping(METRICS_URL)
-    @CrossOrigin
-    fun getMetrics(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam reference: String?,
-    ): CorpusMetrics {
-        logger.info("Get metrics for reference layer $reference and hypothesis layer $job in $corpus")
-
-        val cm = CorpusMetrics(
-            corpora.getReadAccessOrThrow(corpus, request),
-            settings = METRIC_TYPES,
-            hypothesis = job,
-            reference = if (reference.isNullOrBlank()) SOURCE_LAYER_NAME else reference
-        )
-        return cm
-    }
-
-    @GetMapping(METRICS_CSV_URL)
-    @CrossOrigin
-    fun getMetricsSamples(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam reference: String,
-        @RequestParam setting: String,
-        @RequestParam("class") classType: String,
-        @RequestParam group: String?
-    ): ByteArray {
-        val setting = METRIC_TYPES.first { it.id == setting }
-
-        val layerFilter: MetricsLayerFilter? = getLayerFilter(group, setting)
-
-        val cm = CorpusMetrics(
-            corpus = corpora.getReadAccessOrThrow(corpus, request),
-            hypothesis = job,
-            reference = reference,
-            layerFilter = layerFilter,
-            truncate = false,
-            settings = listOf(setting),
-        )
-        val mt = cm.metricTypes.values.first()
-
-        if (group != null) {
-            val fileName = "metrics-$reference-$job-${classType}-${group}.csv"
-            val csv = mt.samplesToCsv(group, classType)
-            return samplesToZip(corpus, job, reference, csv, fileName)
-        } else {
-            val fileName = "metrics-$reference-$job-${mt.setting.id}-${classType}.csv"
-            val csv = mt.samplesToCsv(classType)
-            return samplesToZip(corpus, job, reference, csv, fileName)
-        }
-    }
-
-    private fun getLayerFilter(group: String?, setting: MetricsSettings): MetricsLayerFilter? {
-        val layerFilter: MetricsLayerFilter?
-        if (group != null) {
-            val hypoFilter: PosLemmaTermFilter
-            val refFilter: PosLemmaTermFilter
-            if (setting is PosByPosMetricsSettings) {
-                hypoFilter = PosLemmaTermFilter(group, null)
-                refFilter = PosLemmaTermFilter(group, null)
-            } else {
-                hypoFilter = PosLemmaTermFilter(null, group)
-                refFilter = PosLemmaTermFilter(null, group)
-            }
-            layerFilter = MetricsLayerFilter(hypoFilter, refFilter)
-        } else {
-            layerFilter = null
-        }
-        return layerFilter
-    }
-
-    fun samplesToZip(
-        corpus: UUID,
-        job: String,
-        reference: String?,
-        csvBody: String,
-        fileName: String
-    ): ByteArray {
-        // Create csv file.
-        val dir: File = createTempDirectory("samples").toFile()
-        val file = CSVFile(dir.resolve(fileName))
-        file.appendText(CSVFile.toCSVHeader(listOf("token","$reference lemma","$reference pos","$job lemma","$job pos")))
-        file.appendText(csvBody)
-        // Write metadata & create zip
-        val metadata = writeMetadataToDir(corpus, job, reference, dir)
-        val zipFile = createZipFile(dir.listFiles()!!.asSequence())
-        // Configure response for zip.
-        response!!.contentType = "application/zip"
-        response.setContentDisposition(metadata.name + "-evaluation.zip")
-        // zip the directory
-        return zipFile.readBytes()
-    }
-
-    @GetMapping(CONFUSION_URL)
-    @CrossOrigin
-    fun getConfusion(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam reference: String?,
-    ): Confusion {
-        logger.info("Get confusion for reference layer $reference and hypothesis layer $job in $corpus")
-        return CorpusConfusion(
-            corpora.getReadAccessOrThrow(corpus, request),
-            hypothesis = job,
-            reference = if (reference.isNullOrBlank()) SOURCE_LAYER_NAME else reference,
-        )
-    }
-
-    @GetMapping(CONFUSION_CSV_URL)
-    @CrossOrigin
-    fun getConfusionSamples(
-        @PathVariable corpus: UUID,
-        @PathVariable  job: String,
-        @RequestParam reference: String,
-        @RequestParam hypoPosFilter: String?,
-        @RequestParam refPosFilter: String?,
-    ): ByteArray {
-        val cc = CorpusConfusion(
-            corpus = corpora.getReadAccessOrThrow(corpus, request),
-            hypothesis = job,
-            reference = reference,
-            layerFilter = ConfusionLayerFilter(
-                PosLemmaTermFilter(hypoPosFilter, null),
-                PosLemmaTermFilter(refPosFilter, null)
-            )
-        )
-        val fileName = "confusion-${refPosFilter}-${hypoPosFilter}.csv"
-        val csv = cc.samplesToCSV()
-        return samplesToZip(corpus, job, reference, csv, fileName)
-    }
-
-    @GetMapping(EVALUATION_CSV_URL)
-    @ResponseBody
-    @CrossOrigin
-    fun download(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-        @RequestParam reference: String?,
-        @RequestParam hypothesisPos: String?,
-        @RequestParam referencePos: String?,
-    ): ByteArray {
-        if (reference != null && hypothesisPos != null && referencePos != null)
-            return getConfusionSamples(corpus, job, reference, hypothesisPos, referencePos)
-
-        return executeAndLogTime("GetEvaluationCSVs") {
-            val dir: File = createTempDirectory("evaluation").toFile()
-            createDistributionCsv(dir, corpus, job)
-            if (reference != null) {
-                createMetricsCsv(dir, corpus, job, reference)
-                createConfusionCsv(dir, corpus, job, reference)
-            }
-            val metadata = writeMetadataToDir(corpus, job, reference, dir)
-            response!!.contentType = "application/zip"
-            response.setContentDisposition(metadata.name + "-evaluation.zip")
-
-            // zip the directory
-            val zipFile = createZipFile(dir.listFiles()!!.asSequence())
-            zipFile.readBytes()
-        }
-    }
-
-    private fun createConfusionCsv(dir: File, corpus: UUID, job: String, reference: String?) {
-        val file = CSVFile(dir.resolve("confusion.csv"))
-        file.appendText(getConfusion(corpus, job, reference).countsToCSV())
-    }
-
-    private fun createMetricsCsv(dir: File, corpus: UUID, job: String, reference: String?) {
-        val metrics = getMetrics(corpus, job = job, reference = reference)
-        val globFile = CSVFile(dir.resolve("metrics-global.csv"))
-        globFile.appendText(metrics.toGlobalCsv())
-
-        metrics.metricTypes.values.forEach { mt ->
-            val file = CSVFile(dir.resolve("metrics-${mt.setting.id}.csv"))
-            file.appendText(mt.toGroupedCsv())
-        }
-    }
-
-    private fun createDistributionCsv(dir: File, corpus: UUID, job: String) {
-        val file = CSVFile(dir.resolve("distribution.csv"))
-        file.appendText(getDistribution(corpus, job).toCSV())
-    }
-
-    private fun writeMetadataToDir(
-        corpus: UUID, job: String, reference: String?, dir: File,
-    ): CorpusMetadata {
-        val corpus = corpora.getReadAccessOrThrow(corpus, request)
-        val metadata = corpus.metadata.expensiveGet()
-
-        val metadataFile = File(dir.resolve("metadata.txt").toURI())
-        metadataFile.appendText("Evaluation generated by Galahad\n")
-        metadataFile.appendText("${LocalDateTime.now().format(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"))}\n")
-        metadataFile.appendText("Corpus: ${metadata.name}\n")
-        metadataFile.appendText("Documents: ${metadata.numDocs}\n")
-        metadataFile.appendText("Era: ${metadata.eraFrom}-${metadata.eraTo}\n")
-        metadataFile.appendText("Hypothesis: $job\n")
-        if (reference != null) metadataFile.appendText("Reference: $reference\n")
-        return metadata
-    }
+package org.ivdnt.galahad.evaluation
+
+import jakarta.servlet.http.HttpServletRequest
+import jakarta.servlet.http.HttpServletResponse
+import org.apache.logging.log4j.kotlin.Logging
+import org.ivdnt.galahad.app.*
+import org.ivdnt.galahad.data.CorporaController
+import org.ivdnt.galahad.data.corpus.CorpusMetadata
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
+import org.ivdnt.galahad.evaluation.comparison.ConfusionLayerFilter
+import org.ivdnt.galahad.evaluation.comparison.MetricsLayerFilter
+import org.ivdnt.galahad.evaluation.comparison.PosLemmaTermFilter
+import org.ivdnt.galahad.evaluation.confusion.Confusion
+import org.ivdnt.galahad.evaluation.confusion.CorpusConfusion
+import org.ivdnt.galahad.evaluation.distribution.CorpusDistribution
+import org.ivdnt.galahad.evaluation.metrics.CorpusMetrics
+import org.ivdnt.galahad.evaluation.metrics.METRIC_TYPES
+import org.ivdnt.galahad.evaluation.metrics.MetricsSettings
+import org.ivdnt.galahad.evaluation.metrics.PosByPosMetricsSettings
+import org.ivdnt.galahad.port.csv.CSVFile
+import org.ivdnt.galahad.util.createZipFile
+import org.ivdnt.galahad.util.setContentDisposition
+import org.springframework.beans.factory.annotation.Autowired
+import org.springframework.web.bind.annotation.*
+import java.io.File
+import java.time.LocalDateTime
+import java.time.format.DateTimeFormatter
+import java.util.*
+import kotlin.io.path.createTempDirectory
+
+const val DISTRIBUTION_MAX_SIZE = 1000
+
+@RestController
+class EvaluationController(
+    val corpora: CorporaController,
+) : Logging {
+
+    @Autowired
+    private val request: HttpServletRequest? = null
+
+    @Autowired
+    private val response: HttpServletResponse? = null
+
+    @GetMapping(DISTRIBUTION_URL)
+    @CrossOrigin
+    fun getDistribution(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+    ): CorpusDistribution {
+        logger.info("Get distribution for hypothesis layer $job in $corpus")
+        return CorpusDistribution(
+            corpora.getReadAccessOrThrow(corpus, request),
+            job
+        ).trim(DISTRIBUTION_MAX_SIZE) as CorpusDistribution
+    }
+
+    @GetMapping(METRICS_URL)
+    @CrossOrigin
+    fun getMetrics(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam reference: String?,
+    ): CorpusMetrics {
+        logger.info("Get metrics for reference layer $reference and hypothesis layer $job in $corpus")
+
+        val cm = CorpusMetrics(
+            corpora.getReadAccessOrThrow(corpus, request),
+            settings = METRIC_TYPES,
+            hypothesis = job,
+            reference = if (reference.isNullOrBlank()) SOURCE_LAYER_NAME else reference
+        )
+        return cm
+    }
+
+    @GetMapping(METRICS_CSV_URL)
+    @CrossOrigin
+    fun getMetricsSamples(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam reference: String,
+        @RequestParam setting: String,
+        @RequestParam("class") classType: String,
+        @RequestParam group: String?
+    ): ByteArray {
+        val setting = METRIC_TYPES.first { it.id == setting }
+
+        val layerFilter: MetricsLayerFilter? = getLayerFilter(group, setting)
+
+        val cm = CorpusMetrics(
+            corpus = corpora.getReadAccessOrThrow(corpus, request),
+            hypothesis = job,
+            reference = reference,
+            layerFilter = layerFilter,
+            truncate = false,
+            settings = listOf(setting),
+        )
+        val mt = cm.metricTypes.values.first()
+
+        if (group != null) {
+            val fileName = "metrics-$reference-$job-${classType}-${group}.csv"
+            val csv = mt.samplesToCsv(group, classType)
+            return samplesToZip(corpus, job, reference, csv, fileName)
+        } else {
+            val fileName = "metrics-$reference-$job-${mt.setting.id}-${classType}.csv"
+            val csv = mt.samplesToCsv(classType)
+            return samplesToZip(corpus, job, reference, csv, fileName)
+        }
+    }
+
+    private fun getLayerFilter(group: String?, setting: MetricsSettings): MetricsLayerFilter? {
+        val layerFilter: MetricsLayerFilter?
+        if (group != null) {
+            val hypoFilter: PosLemmaTermFilter
+            val refFilter: PosLemmaTermFilter
+            if (setting is PosByPosMetricsSettings) {
+                hypoFilter = PosLemmaTermFilter(group, null)
+                refFilter = PosLemmaTermFilter(group, null)
+            } else {
+                hypoFilter = PosLemmaTermFilter(null, group)
+                refFilter = PosLemmaTermFilter(null, group)
+            }
+            layerFilter = MetricsLayerFilter(hypoFilter, refFilter)
+        } else {
+            layerFilter = null
+        }
+        return layerFilter
+    }
+
+    fun samplesToZip(
+        corpus: UUID,
+        job: String,
+        reference: String?,
+        csvBody: String,
+        fileName: String
+    ): ByteArray {
+        // Create csv file.
+        val dir: File = createTempDirectory("samples").toFile()
+        val file = CSVFile(dir.resolve(fileName))
+        file.appendText(CSVFile.toCSVHeader(listOf("token","$reference lemma","$reference pos","$job lemma","$job pos")))
+        file.appendText(csvBody)
+        // Write metadata & create zip
+        val metadata = writeMetadataToDir(corpus, job, reference, dir)
+        val zipFile = createZipFile(dir.listFiles()!!.asSequence())
+        // Configure response for zip.
+        response!!.contentType = "application/zip"
+        response.setContentDisposition(metadata.name + "-evaluation.zip")
+        // zip the directory
+        return zipFile.readBytes()
+    }
+
+    @GetMapping(CONFUSION_URL)
+    @CrossOrigin
+    fun getConfusion(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam reference: String?,
+    ): Confusion {
+        logger.info("Get confusion for reference layer $reference and hypothesis layer $job in $corpus")
+        return CorpusConfusion(
+            corpora.getReadAccessOrThrow(corpus, request),
+            hypothesis = job,
+            reference = if (reference.isNullOrBlank()) SOURCE_LAYER_NAME else reference,
+        )
+    }
+
+    @GetMapping(CONFUSION_CSV_URL)
+    @CrossOrigin
+    fun getConfusionSamples(
+        @PathVariable corpus: UUID,
+        @PathVariable  job: String,
+        @RequestParam reference: String,
+        @RequestParam hypoPosFilter: String?,
+        @RequestParam refPosFilter: String?,
+    ): ByteArray {
+        val cc = CorpusConfusion(
+            corpus = corpora.getReadAccessOrThrow(corpus, request),
+            hypothesis = job,
+            reference = reference,
+            layerFilter = ConfusionLayerFilter(
+                PosLemmaTermFilter(hypoPosFilter, null),
+                PosLemmaTermFilter(refPosFilter, null)
+            )
+        )
+        val fileName = "confusion-${refPosFilter}-${hypoPosFilter}.csv"
+        val csv = cc.samplesToCSV()
+        return samplesToZip(corpus, job, reference, csv, fileName)
+    }
+
+    @GetMapping(EVALUATION_CSV_URL)
+    @ResponseBody
+    @CrossOrigin
+    fun download(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+        @RequestParam reference: String?,
+        @RequestParam hypothesisPos: String?,
+        @RequestParam referencePos: String?,
+    ): ByteArray {
+        if (reference != null && hypothesisPos != null && referencePos != null)
+            return getConfusionSamples(corpus, job, reference, hypothesisPos, referencePos)
+
+        return executeAndLogTime("GetEvaluationCSVs") {
+            val dir: File = createTempDirectory("evaluation").toFile()
+            createDistributionCsv(dir, corpus, job)
+            if (reference != null) {
+                createMetricsCsv(dir, corpus, job, reference)
+                createConfusionCsv(dir, corpus, job, reference)
+            }
+            val metadata = writeMetadataToDir(corpus, job, reference, dir)
+            response!!.contentType = "application/zip"
+            response.setContentDisposition(metadata.name + "-evaluation.zip")
+
+            // zip the directory
+            val zipFile = createZipFile(dir.listFiles()!!.asSequence())
+            zipFile.readBytes()
+        }
+    }
+
+    private fun createConfusionCsv(dir: File, corpus: UUID, job: String, reference: String?) {
+        val file = CSVFile(dir.resolve("confusion.csv"))
+        file.appendText(getConfusion(corpus, job, reference).countsToCSV())
+    }
+
+    private fun createMetricsCsv(dir: File, corpus: UUID, job: String, reference: String?) {
+        val metrics = getMetrics(corpus, job = job, reference = reference)
+        val globFile = CSVFile(dir.resolve("metrics-global.csv"))
+        globFile.appendText(metrics.toGlobalCsv())
+
+        metrics.metricTypes.values.forEach { mt ->
+            val file = CSVFile(dir.resolve("metrics-${mt.setting.id}.csv"))
+            file.appendText(mt.toGroupedCsv())
+        }
+    }
+
+    private fun createDistributionCsv(dir: File, corpus: UUID, job: String) {
+        val file = CSVFile(dir.resolve("distribution.csv"))
+        file.appendText(getDistribution(corpus, job).toCSV())
+    }
+
+    private fun writeMetadataToDir(
+        corpus: UUID, job: String, reference: String?, dir: File,
+    ): CorpusMetadata {
+        val corpus = corpora.getReadAccessOrThrow(corpus, request)
+        val metadata = corpus.metadata.expensiveGet()
+
+        val metadataFile = File(dir.resolve("metadata.txt").toURI())
+        metadataFile.appendText("Evaluation generated by Galahad\n")
+        metadataFile.appendText("${LocalDateTime.now().format(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"))}\n")
+        metadataFile.appendText("Corpus: ${metadata.name}\n")
+        metadataFile.appendText("Documents: ${metadata.numDocs}\n")
+        metadataFile.appendText("Era: ${metadata.eraFrom}-${metadata.eraTo}\n")
+        metadataFile.appendText("Hypothesis: $job\n")
+        if (reference != null) metadataFile.appendText("Reference: $reference\n")
+        return metadata
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/assays/AssaysController.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/assays/AssaysController.kt
index 4fdfe1e..59a1ba8 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/assays/AssaysController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/assays/AssaysController.kt
@@ -1,106 +1,105 @@
-package org.ivdnt.galahad.evaluation.assays
-
-import jakarta.servlet.http.HttpServletRequest
-import jakarta.servlet.http.HttpServletResponse
-import org.apache.logging.log4j.kotlin.Logging
-import org.ivdnt.galahad.FileBackedCache
-import org.ivdnt.galahad.app.ASSAYS_URL
-import org.ivdnt.galahad.app.ASSAY_URL
-import org.ivdnt.galahad.data.CorporaController
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
-import org.ivdnt.galahad.evaluation.metrics.FlatMetricType
-import org.ivdnt.galahad.evaluation.metrics.FlatMetricTypeAssay
-import org.springframework.beans.factory.annotation.Autowired
-import org.springframework.web.bind.annotation.*
-import java.util.*
-import kotlin.collections.HashMap
-
-/**
- * {
- *     "dataset-1": {
- *         "posByPos": {
- *             "tagger-1": {
- *                 "micro": { ... }, "macro": { ... }
- *             },
- *             "tagger-2": { ... },
- *         },
- *         "lemmaByLemma": { ... },
- *     },
- *     "dataset-2": { ... },
- * }
- */
-typealias AssaysMatrix = Map<String, Map<String, FlatMetricTypeAssay>>
-typealias MutableAssaysMatrix = MutableMap<String, MutableMap<String, MutableMap<String, FlatMetricType>>>
-
-@RestController
-class AssaysController(
-    val corpora: CorporaController,
-) : Logging {
-
-    @Autowired
-    private val request: HttpServletRequest? = null
-
-    @Autowired
-    private val response: HttpServletResponse? = null
-
-    /**
-     * A matrix of 'tagger' -> 'dataset' -> 'FlatMetric' -> 'scores per category',
-     * for all datasets corpora that have been tagged with at least one tagger, excluding the sourceLayer.
-     */
-    val assaysMatrix = object : FileBackedCache<AssaysMatrix>(corpora.assaysFile, HashMap()) {
-        override fun isValid(lastModified: Long): Boolean {
-            return corpora.datasets.firstOrNull { it.lastModified > lastModified } == null
-            TODO("Maybe just check the validity of the other assays?")
-        }
-
-        override fun set(): AssaysMatrix {
-            // tagger -> dataset -> assay
-            val assaysMatrix: MutableAssaysMatrix = HashMap()
-            // For all datasets
-            corpora.datasets.forEach { dataset ->
-                // For all jobs in the dataset
-                dataset.jobs.readAll()
-                    // Skip the source layer
-                    .filter { it.name != SOURCE_LAYER_NAME }
-                    // Add the assay to the matrix
-                    .forEach { job ->
-                        val meta = dataset.metadata.expensiveGet()
-                        // Initialize the dataset row if needed
-                        if (assaysMatrix[meta.name] == null) {
-                            assaysMatrix[meta.name] = HashMap()
-                        }
-                        val assay = getAssay(meta.uuid, job.name)
-                        assay?.forEach {
-                            assaysMatrix[meta.name]?.putIfAbsent(it.key, HashMap())
-                            assaysMatrix[meta.name]?.get(it.key)?.put(job.name, it.value)
-                        }
-                    }
-            }
-            return assaysMatrix
-        }
-    }
-
-    /**
-     * Get the assay for a single job in a specific corpus. Also used to construct [assaysMatrix].
-     */
-    @GetMapping(ASSAY_URL)
-    @CrossOrigin
-    fun getAssay(
-        @PathVariable corpus: UUID,
-        @PathVariable job: String,
-    ): FlatMetricTypeAssay? {
-        // The assay is some simple, preferably plain numerical, value that indicates preformance
-        // Since the exact requirements for the definition of an assays might still change
-        // we don't provide a solid definition, but instead it is defined ad hoc here
-        return corpora.getReadAccessOrThrow(corpus, request).jobs.readOrNull(job)?.assay?.get<FlatMetricTypeAssay>()
-    }
-
-    /**
-     * Return [assaysMatrix].
-     */
-    @RequestMapping(value = [ASSAYS_URL], method = [RequestMethod.GET], produces = ["application/json"])
-    @CrossOrigin
-    fun getAssays(): AssaysMatrix {
-        return assaysMatrix.get<AssaysMatrix>()
-    }
+package org.ivdnt.galahad.evaluation.assays
+
+import jakarta.servlet.http.HttpServletRequest
+import jakarta.servlet.http.HttpServletResponse
+import org.apache.logging.log4j.kotlin.Logging
+import org.ivdnt.galahad.FileBackedCache
+import org.ivdnt.galahad.app.ASSAYS_URL
+import org.ivdnt.galahad.app.ASSAY_URL
+import org.ivdnt.galahad.data.CorporaController
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
+import org.ivdnt.galahad.evaluation.metrics.FlatMetricType
+import org.ivdnt.galahad.evaluation.metrics.FlatMetricTypeAssay
+import org.springframework.beans.factory.annotation.Autowired
+import org.springframework.web.bind.annotation.*
+import java.util.*
+
+/**
+ * {
+ *     "dataset-1": {
+ *         "posByPos": {
+ *             "tagger-1": {
+ *                 "micro": { ... }, "macro": { ... }
+ *             },
+ *             "tagger-2": { ... },
+ *         },
+ *         "lemmaByLemma": { ... },
+ *     },
+ *     "dataset-2": { ... },
+ * }
+ */
+typealias AssaysMatrix = Map<String, Map<String, FlatMetricTypeAssay>>
+typealias MutableAssaysMatrix = MutableMap<String, MutableMap<String, MutableMap<String, FlatMetricType>>>
+
+@RestController
+class AssaysController(
+    val corpora: CorporaController,
+) : Logging {
+
+    @Autowired
+    private val request: HttpServletRequest? = null
+
+    @Autowired
+    private val response: HttpServletResponse? = null
+
+    /**
+     * A matrix of 'tagger' -> 'dataset' -> 'FlatMetric' -> 'scores per category',
+     * for all datasets corpora that have been tagged with at least one tagger, excluding the sourceLayer.
+     */
+    val assaysMatrix = object : FileBackedCache<AssaysMatrix>(corpora.assaysFile, HashMap()) {
+        override fun isValid(lastModified: Long): Boolean {
+            return corpora.datasets.firstOrNull { it.lastModified > lastModified } == null
+            TODO("Maybe just check the validity of the other assays?")
+        }
+
+        override fun set(): AssaysMatrix {
+            // tagger -> dataset -> assay
+            val assaysMatrix: MutableAssaysMatrix = HashMap()
+            // For all datasets
+            corpora.datasets.forEach { dataset ->
+                // For all jobs in the dataset
+                dataset.jobs.readAll()
+                    // Skip the source layer
+                    .filter { it.name != SOURCE_LAYER_NAME }
+                    // Add the assay to the matrix
+                    .forEach { job ->
+                        val meta = dataset.metadata.expensiveGet()
+                        // Initialize the dataset row if needed
+                        if (assaysMatrix[meta.name] == null) {
+                            assaysMatrix[meta.name] = HashMap()
+                        }
+                        val assay = getAssay(meta.uuid, job.name)
+                        assay?.forEach {
+                            assaysMatrix[meta.name]?.putIfAbsent(it.key, HashMap())
+                            assaysMatrix[meta.name]?.get(it.key)?.put(job.name, it.value)
+                        }
+                    }
+            }
+            return assaysMatrix
+        }
+    }
+
+    /**
+     * Get the assay for a single job in a specific corpus. Also used to construct [assaysMatrix].
+     */
+    @GetMapping(ASSAY_URL)
+    @CrossOrigin
+    fun getAssay(
+        @PathVariable corpus: UUID,
+        @PathVariable job: String,
+    ): FlatMetricTypeAssay? {
+        // The assay is some simple, preferably plain numerical, value that indicates preformance
+        // Since the exact requirements for the definition of an assays might still change
+        // we don't provide a solid definition, but instead it is defined ad hoc here
+        return corpora.getReadAccessOrThrow(corpus, request).jobs.readOrNull(job)?.assay?.get<FlatMetricTypeAssay>()
+    }
+
+    /**
+     * Return [assaysMatrix].
+     */
+    @RequestMapping(value = [ASSAYS_URL], method = [RequestMethod.GET], produces = ["application/json"])
+    @CrossOrigin
+    fun getAssays(): AssaysMatrix {
+        return assaysMatrix.get<AssaysMatrix>()
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/LayerComparison.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/LayerComparison.kt
index e02745c..b0b174c 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/LayerComparison.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/LayerComparison.kt
@@ -1,151 +1,156 @@
-package org.ivdnt.galahad.evaluation.comparison
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.data.layer.Term
-
-fun ListIterator<Term>.nextOrNull(): Term? {
-    val iter = iterator()
-    return if (iter.hasNext()) iter.next() else null
-}
-
-// Some hardcoded punctuation
-val PUNCTUATION = listOf(",", ".", "?", "!", ":", ";", ")", "(", "'", "\"")
-
-/**
- * Match the [Layer.terms] of two layers based on their position (offset and length), not on their actual value (literal/pos/lemma).
- * When pos filters are provide, only match [TermComparison] of that pos.
- */
-class LayerComparison(
-    private val hypothesisLayer: Layer,
-    private val referenceLayer: Layer,
-    private val layerFilter: LayerFilter? = null,
-) {
-    @JsonIgnore
-    val matches: MutableList<TermComparison> = ArrayList()
-
-    @JsonIgnore
-    val referenceTermsWithoutMatches: MutableList<Term> = ArrayList()
-
-    @JsonIgnore
-    val hypothesisTermsWithoutMatches: MutableList<Term> = ArrayList()
-    @JsonIgnore
-    private val hypoIter: ListIterator<Term> = iterForTermsInLayer(hypothesisLayer)
-    @JsonIgnore
-    private val refIter: ListIterator<Term> = iterForTermsInLayer(referenceLayer)
-    @JsonIgnore
-    private var currentHypoTerm: Term? = Term.EMPTY
-    @JsonIgnore
-    private var currentRefTerm: Term? = Term.EMPTY
-
-    init {
-        if (refIter.hasNext() && hypoIter.hasNext()) {
-            compare()
-        } else {
-            hypothesisTermsWithoutMatches.addAll(hypothesisLayer.terms)
-            referenceTermsWithoutMatches.addAll(referenceLayer.terms)
-        }
-    }
-
-    /** Iterate through the terms of both layers simultaneously and compare them. */
-    private fun compare() {
-        // First terms
-        nextHypo()
-        nextRef()
-        // While there are next terms
-        while (currentHypoTerm != null && currentRefTerm != null) {
-            val comp = TermComparison(hypoTerm = currentHypoTerm!!, refTerm = currentRefTerm!!)
-            compareTerm(comp)
-        }
-        // One of the two could be non-null. These are not included in the remaining refIter.
-        currentHypoTerm?.let(::hypoNoMatch)
-        currentRefTerm?.let(::refNoMatch)
-        // The remaining terms have no matches
-        hypoIter.forEachRemaining(::hypoNoMatch)
-        refIter.forEachRemaining(::refNoMatch)
-    }
-
-    private fun compareTerm(comp: TermComparison) {
-        // Act on the comparison
-        if (comp.fullOverlap) {
-            fullMatch(comp)
-        } else {
-            // Unequal first offset
-            if (comp.hypoTerm.firstOffset < comp.refTerm.firstOffset) {
-                hypoNoMatch()
-            } else if (comp.hypoTerm.firstOffset > comp.refTerm.firstOffset) {
-                refNoMatch()
-            }
-            // Equal first offset but no match.
-            // Try to truncate either terms to see if the last char is punctuation.
-            else if (symmetricTruncatedPcMatch(comp)) {
-                // If so, still match it.
-                fullMatch(comp)
-            } else {
-                hypoNoMatch()
-                refNoMatch()
-            }
-        }
-    }
-
-    private fun fullMatch(termComparison: TermComparison) {
-        if (layerFilter?.filter(termComparison) != false) {
-            matches.add(termComparison)
-        }
-        nextHypo()
-        nextRef()
-    }
-
-    private fun hypoNoMatch() {
-        hypoNoMatch(currentHypoTerm!!)
-        nextHypo()
-    }
-
-    private fun hypoNoMatch(t: Term) {
-        // Note how layerFilter can be null, and both null and true != false.
-        if (layerFilter?.hypoTermFilter?.filter(t) != false) {
-            hypothesisTermsWithoutMatches.add(t)
-        }
-    }
-
-    private fun refNoMatch() {
-        refNoMatch(currentRefTerm!!)
-        nextRef()
-    }
-
-    private fun refNoMatch(t: Term) {
-        if (layerFilter?.refTermFilter?.filter(t) != false) {
-            referenceTermsWithoutMatches.add(t)
-        }
-    }
-
-    private fun nextHypo() {
-        currentHypoTerm = hypoIter.nextOrNull()
-    }
-
-    private fun nextRef() {
-        currentRefTerm = refIter.nextOrNull()
-    }
-
-    /** Iterate through the terms of the layer sorted on offset. */
-    private fun iterForTermsInLayer(layer: Layer): ListIterator<Term> {
-        return layer.terms
-            // Terms can only be a match if their first offset is the same
-            .sortedBy { it.firstOffset }.listIterator()
-    }
-
-    private fun symmetricTruncatedPcMatch(comp: TermComparison): Boolean {
-        val aStr: String = comp.hypoTerm.literals
-        val bStr: String = comp.refTerm.literals
-        return truncatedPcMatch(aStr, bStr) || truncatedPcMatch(bStr, aStr)
-    }
-
-    private fun truncatedPcMatch(aStr: String, bStr: String): Boolean {
-        if (PUNCTUATION.contains(aStr.last().toString())) {
-            if (aStr.slice(0 until aStr.lastIndex) == bStr) {
-                return true
-            }
-        }
-        return false
-    }
+package org.ivdnt.galahad.evaluation.comparison
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.data.layer.Term
+
+fun ListIterator<Term>.nextOrNull(): Term? {
+    val iter = iterator()
+    return if (iter.hasNext()) iter.next() else null
+}
+
+// Some hardcoded punctuation
+val PUNCTUATION = listOf(",", ".", "?", "!", ":", ";", ")", "(", "'", "\"")
+
+/**
+ * Match the [Layer.terms] of two layers based on their position (offset and length), not on their actual value (literal/pos/lemma).
+ * When pos filters are provide, only match [TermComparison] of that pos.
+ */
+class LayerComparison(
+    private val hypothesisLayer: Layer,
+    private val referenceLayer: Layer,
+    private val layerFilter: LayerFilter? = null,
+) {
+    @JsonIgnore
+    val matches: MutableList<TermComparison> = ArrayList()
+
+    @JsonIgnore
+    val referenceTermsWithoutMatches: MutableList<Term> = ArrayList()
+
+    @JsonIgnore
+    val hypothesisTermsWithoutMatches: MutableList<Term> = ArrayList()
+    @JsonIgnore
+    private val hypoIter: ListIterator<Term> = iterForTermsInLayer(hypothesisLayer)
+    @JsonIgnore
+    private val refIter: ListIterator<Term> = iterForTermsInLayer(referenceLayer)
+    @JsonIgnore
+    private var currentHypoTerm: Term? = Term.EMPTY
+    @JsonIgnore
+    private var currentRefTerm: Term? = Term.EMPTY
+
+    init {
+        if (refIter.hasNext() && hypoIter.hasNext()) {
+            compare()
+        } else {
+            hypothesisTermsWithoutMatches.addAll(hypothesisLayer.terms)
+            referenceTermsWithoutMatches.addAll(referenceLayer.terms)
+        }
+    }
+
+    /** Iterate through the terms of both layers simultaneously and compare them. */
+    private fun compare() {
+        // First terms
+        nextHypo()
+        nextRef()
+        // While there are next terms
+        while (currentHypoTerm != null && currentRefTerm != null) {
+            val comp = TermComparison(hypoTerm = currentHypoTerm!!, refTerm = currentRefTerm!!)
+            compareTerm(comp)
+        }
+        // One of the two could be non-null. These are not included in the remaining refIter.
+        currentHypoTerm?.let(::hypoNoMatch)
+        currentRefTerm?.let(::refNoMatch)
+        // The remaining terms have no matches
+        hypoIter.forEachRemaining(::hypoNoMatch)
+        refIter.forEachRemaining(::refNoMatch)
+    }
+
+    private fun compareTerm(comp: TermComparison) {
+        // Act on the comparison
+        if (comp.fullOverlap) {
+            fullMatch(comp)
+        } else {
+            // Unequal first offset
+            if (comp.hypoTerm.firstOffset < comp.refTerm.firstOffset) {
+                hypoNoMatch()
+            } else if (comp.hypoTerm.firstOffset > comp.refTerm.firstOffset) {
+                refNoMatch()
+            }
+            // Equal first offset but no match.
+            // Try to truncate either terms to see if the last char is punctuation.
+            else if (symmetricTruncatedPcMatch(comp)) {
+                // If so, still match it.
+                fullMatch(comp)
+            } else {
+                hypoNoMatch()
+                refNoMatch()
+            }
+        }
+    }
+
+    private fun fullMatch(termComparison: TermComparison) {
+        if (layerFilter?.filter(termComparison) != false) {
+            matches.add(termComparison)
+        }
+        nextHypo()
+        nextRef()
+    }
+
+    private fun hypoNoMatch() {
+        hypoNoMatch(currentHypoTerm!!)
+        nextHypo()
+    }
+
+    private fun hypoNoMatch(t: Term) {
+        // Note how layerFilter can be null, and both null and true != false.
+        if (layerFilter?.hypoTermFilter?.filter(t) != false) {
+            hypothesisTermsWithoutMatches.add(t)
+        }
+    }
+
+    private fun refNoMatch() {
+        refNoMatch(currentRefTerm!!)
+        nextRef()
+    }
+
+    private fun refNoMatch(t: Term) {
+        if (layerFilter?.refTermFilter?.filter(t) != false) {
+            referenceTermsWithoutMatches.add(t)
+        }
+    }
+
+    private fun nextHypo() {
+        currentHypoTerm = hypoIter.nextOrNull()
+    }
+
+    private fun nextRef() {
+        currentRefTerm = refIter.nextOrNull()
+    }
+
+    /** Iterate through the terms of the layer sorted on offset. */
+    private fun iterForTermsInLayer(layer: Layer): ListIterator<Term> {
+        return layer.terms
+            // Terms can only be a match if their first offset is the same
+            .sortedBy { it.firstOffset }.listIterator()
+    }
+
+    private fun symmetricTruncatedPcMatch(comp: TermComparison): Boolean {
+        val aStr: String = comp.hypoTerm.literals
+        val bStr: String = comp.refTerm.literals
+        return truncatedPcMatch(aStr, bStr) || truncatedPcMatch(bStr, aStr)
+    }
+
+    companion object {
+        fun truncatedPcMatch(aStr: String, bStr: String): Boolean {
+            if (aStr.isEmpty() || bStr.isEmpty()) {
+                return false
+            }
+            if (PUNCTUATION.contains(aStr.last().toString())) {
+                if (aStr.slice(0 until aStr.lastIndex) == bStr) {
+                    return true
+                }
+            }
+            return false
+        }
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermComparison.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermComparison.kt
index 9811ec7..534ee40 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermComparison.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermComparison.kt
@@ -1,93 +1,93 @@
-package org.ivdnt.galahad.evaluation.comparison
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import org.ivdnt.galahad.data.layer.Term
-import org.ivdnt.galahad.data.layer.WordForm
-
-fun <T> symmetricDifference(
-    set1: Set<T>,
-    set2: Set<T>,
-    equals: (T, T) -> Boolean,
-): Set<T> {
-
-    val mset1: MutableSet<T> = HashSet(set1)
-    val mset2: MutableSet<T> = HashSet(set2)
-
-    for (v1 in set1) {
-        for (v2 in set2) {
-            if (equals(v1, v2)) {
-                mset1.remove(v1)
-                mset2.remove(v2)
-            }
-        }
-    }
-
-    return mset1 union mset2
-}
-
-data class TermComparison(
-    val hypoTerm: Term, // Hypothesis
-    val refTerm: Term, // True reference
-) {
-    /** Full overlap dependent on the word forms. Overlap of position, not lemma/pos. */
-    @get:JsonIgnore
-    val fullOverlap: Boolean
-        get() = symmetricDifference(hypoTerm.targets.toSet(), refTerm.targets.toSet(),
-                                    equals = { wf1: WordForm, wf2: WordForm ->
-                WordFormComparison(wf1, wf2).fullOverlap
-            }).isEmpty()
-
-    /** Partial overlap dependent on the word forms. Overlap of position, not lemma/pos. */
-    // Currently not used.
-    @get:JsonIgnore
-    val partialOverlap: Boolean
-        get() {
-            hypoTerm.targets.forEach { target1 ->
-                refTerm.targets.forEach { target2 ->
-                    if (WordFormComparison(target1, target2).partialOverlap) {
-                        return true
-                    }
-                }
-            }
-            return false
-        }
-
-    @get:JsonIgnore
-    val equalPosLemma: Boolean
-        get() {
-            return (equalLemma) && (equalPOS)
-        }
-
-    /** Whether the lemma is equal. When the reference lemma is empty or null, any hypothesis lemma is fine. */
-    @get:JsonIgnore
-    val equalLemma: Boolean
-        get() {
-            if (refTerm.lemma == null) return true
-            if (refTerm.lemma!!.isEmpty()) return true
-            if (hypoTerm.lemma == null) return false
-            return hypoTerm.lemma!!.equals(refTerm.lemma!!, true)
-        }
-
-    /** Whether the pos is equal. When the reference pos is empty or null, any hypothesis pos is fine. */
-    @get:JsonIgnore
-    val equalPOS: Boolean
-        get() {
-            if (refTerm.pos == null) return true
-            if (refTerm.pos!!.isEmpty()) return true
-            if (hypoTerm.pos == null) return false
-            return hypoTerm.pos.equals(refTerm.pos, true)
-        }
-
-    @get:JsonIgnore
-    val equalGroupPosHead: Boolean
-        get() {
-            if (refTerm.posHeadGroup == null) return true
-            if (refTerm.posHeadGroup!!.isEmpty()) return true
-            if (hypoTerm.posHeadGroup == null) return false
-            return hypoTerm.posHeadGroup.equals(refTerm.posHeadGroup, true)
-        }
-
-    companion object {
-        const val MISSING_MATCH = "Missing match"
-    }
+package org.ivdnt.galahad.evaluation.comparison
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.data.layer.WordForm
+
+fun <T> symmetricDifference(
+    set1: Set<T>,
+    set2: Set<T>,
+    equals: (T, T) -> Boolean,
+): Set<T> {
+
+    val mset1: MutableSet<T> = HashSet(set1)
+    val mset2: MutableSet<T> = HashSet(set2)
+
+    for (v1 in set1) {
+        for (v2 in set2) {
+            if (equals(v1, v2)) {
+                mset1.remove(v1)
+                mset2.remove(v2)
+            }
+        }
+    }
+
+    return mset1 union mset2
+}
+
+data class TermComparison(
+    val hypoTerm: Term, // Hypothesis
+    val refTerm: Term, // True reference
+) {
+    /** Full overlap dependent on the word forms. Overlap of position, not lemma/pos. */
+    @get:JsonIgnore
+    val fullOverlap: Boolean
+        get() = symmetricDifference(hypoTerm.targets.toSet(), refTerm.targets.toSet(),
+                                    equals = { wf1: WordForm, wf2: WordForm ->
+                WordFormComparison(wf1, wf2).fullOverlap
+            }).isEmpty()
+
+    /** Partial overlap dependent on the word forms. Overlap of position, not lemma/pos. */
+    // Currently not used.
+    @get:JsonIgnore
+    val partialOverlap: Boolean
+        get() {
+            hypoTerm.targets.forEach { target1 ->
+                refTerm.targets.forEach { target2 ->
+                    if (WordFormComparison(target1, target2).partialOverlap) {
+                        return true
+                    }
+                }
+            }
+            return false
+        }
+
+    @get:JsonIgnore
+    val equalPosLemma: Boolean
+        get() {
+            return (equalLemma) && (equalPOS)
+        }
+
+    /** Whether the lemma is equal. When the reference lemma is empty or null, any hypothesis lemma is fine. */
+    @get:JsonIgnore
+    val equalLemma: Boolean
+        get() {
+            if (refTerm.lemma == null) return true
+            if (refTerm.lemma.isEmpty()) return true
+            if (hypoTerm.lemma == null) return false
+            return hypoTerm.lemma.equals(refTerm.lemma, true)
+        }
+
+    /** Whether the pos is equal. When the reference pos is empty or null, any hypothesis pos is fine. */
+    @get:JsonIgnore
+    val equalPOS: Boolean
+        get() {
+            if (refTerm.pos == null) return true
+            if (refTerm.pos.isEmpty()) return true
+            if (hypoTerm.pos == null) return false
+            return hypoTerm.pos.equals(refTerm.pos, true)
+        }
+
+    @get:JsonIgnore
+    val equalGroupPosHead: Boolean
+        get() {
+            if (refTerm.posHeadGroup == null) return true
+            if (refTerm.posHeadGroup!!.isEmpty()) return true
+            if (hypoTerm.posHeadGroup == null) return false
+            return hypoTerm.posHeadGroup.equals(refTerm.posHeadGroup, true)
+        }
+
+    companion object {
+        const val MISSING_MATCH = "Missing match"
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermFilter.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermFilter.kt
index c0b729e..4bf7265 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermFilter.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/comparison/TermFilter.kt
@@ -13,9 +13,9 @@ class PosLemmaTermFilter (
     val posHeadGroup: String? = null,
     val lemma: String? = null,
 ) : TermFilter {
-    val multiplePosFilter: (Term) -> Boolean = { t: Term -> t.isMultiPos }
-    val otherPosFilter: (Term) -> Boolean = { t: Term -> t.pos?.matches(OTHER_POS_REGEX.toRegex()) ?: false}
-    val singlePosFilter: (Term) -> Boolean = { t: Term -> t.posHeadGroupOrDefault == posHeadGroup }
+    private val multiplePosFilter: (Term) -> Boolean = { t: Term -> t.isMultiPos }
+    private val otherPosFilter: (Term) -> Boolean = { t: Term -> t.pos?.contains(Regex(OTHER_POS_REGEX)) ?: false}
+    private val singlePosFilter: (Term) -> Boolean = { t: Term -> t.posHeadGroupOrDefault == posHeadGroup }
     val posFilter: (Term) -> Boolean
     val lemmaFilter: (Term) -> Boolean
 
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/confusion/Confusion.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/confusion/Confusion.kt
index 256c353..fab0380 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/confusion/Confusion.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/confusion/Confusion.kt
@@ -9,7 +9,7 @@ import org.ivdnt.galahad.port.csv.CSVFile
 
 const val MULTIPLE_POS = "MULTIPLE"
 const val OTHER_POS = "OTHER"
-const val OTHER_POS_REGEX = """^[^A-Z]"""
+const val OTHER_POS_REGEX = """^[^a-zA-Z]"""
 
 /**
  * Generic class for the part of speech confusion of a corpus or document.
@@ -111,8 +111,8 @@ open class Confusion(private val truncate: Boolean = true): CsvSampleExporter {
             pos1.contains('+') -> add(MULTIPLE_POS, pos2, evaluationEntry)
             pos2.contains('+') -> add(pos1, MULTIPLE_POS, evaluationEntry)
             // Non-alphabetical pos are mapped to a single category "other"
-            pos1.matches(OTHER_POS_REGEX.toRegex()) -> add(OTHER_POS, pos2, evaluationEntry)
-            pos2.matches(OTHER_POS_REGEX.toRegex()) -> add(pos1, OTHER_POS, evaluationEntry)
+            pos1.contains(Regex(OTHER_POS_REGEX)) -> add(OTHER_POS, pos2, evaluationEntry)
+            pos2.contains(Regex(OTHER_POS_REGEX)) -> add(pos1, OTHER_POS, evaluationEntry)
             // Otherwise a simple merge
             else -> matrix.merge(Pair(pos1, pos2), evaluationEntry) { a, b -> EvaluationEntry.add(a, b, truncate) }
         }
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/distribution/Distribution.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/distribution/Distribution.kt
index 390912b..9a959bb 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/distribution/Distribution.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/distribution/Distribution.kt
@@ -1,90 +1,89 @@
-package org.ivdnt.galahad.evaluation.distribution
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import org.ivdnt.galahad.data.layer.Term
-
-/**
- * Generic class for frequency distributions of terms in a corpus or document.
- * The idea is to sum up the distribution as we go through the terms one by one using [add].
- */
-open class Distribution {
-
-    /**
-     * (lem, pos) -> (count, literal[])
-     */
-    @JsonIgnore
-    val distributionMap: MutableMap<Pair<String, String>, Pair<Int, LiteralsEntry>> = HashMap()
-
-    var isTrimmed = false
-    var coveredChars = 0
-    var coveredAlphabeticChars = 0
-    var totalChars = 0
-    var totalAlphabeticChars = 0
-
-    /**
-     * Is serialized and send through API, so it is in fact used.
-     */
-    @Suppress("unused")
-    val distribution: Set<DistributionRow>
-        get() = distributionMap.entries.map {
-                DistributionRow(
-                    it.key.first,
-                    it.key.second,
-                    it.value.first,
-                    it.value.second
-                )
-            }.toSet()
-
-    fun add(term: Term) {
-        val literal: String = term.literals
-        coveredChars += literal.length
-        coveredAlphabeticChars += literal.count { char -> char.isLetter() }
-        add(
-            lemma = term.lemma ?: Term.NO_LEMMA,
-            pos = term.posHeadGroup ?: Term.NO_POS,
-            count = 1,
-            literals = LiteralsEntry(mapOf(term.literals to 1))
-        )
-    }
-
-    private fun add(lemma: String, pos: String, count: Int, literals: LiteralsEntry) {
-        distributionMap.merge(
-            Pair(lemma, pos), Pair(count, literals)
-        ) { p1, p2 -> Pair(p1.first + p2.first, p1.second.add(p2.second)) }
-    }
-
-    fun add(other: Distribution) {
-        coveredChars += other.coveredChars
-        coveredAlphabeticChars += other.coveredAlphabeticChars
-        totalChars += other.totalChars
-        totalAlphabeticChars += other.totalAlphabeticChars
-        other.distributionMap.forEach {
-            add(it.key.first, it.key.second, it.value.first, it.value.second)
-        }
-    }
-
-    fun toCSV(): String {
-        var csv: String = DistributionRow.getCsvHeader()
-        getSorted().forEach { csv += it.toCSVRecord() }
-        return csv
-    }
-
-    @JsonIgnore
-    fun getSorted(): Set<DistributionRow> {
-        return distributionMap.toList()
-            // Sort on count (descending, hence minus), then on lemma.
-            // Note that 'it' is of type Pair< Pair<String, String> , Pair<Int, LiteralsEntry> >
-            .sortedWith(compareBy({ -it.second.first }, { it.first.first.lowercase() })).toMap()
-            // Map to serializable rows.
-            .entries.map { DistributionRow(it.key.first, it.key.second, it.value.first, it.value.second) }.toSet()
-    }
-
-    fun trim(maxSize: Int, minVal: Int = 2): Distribution {
-        if (distributionMap.size > maxSize) {
-            isTrimmed = true
-            distributionMap.entries.removeIf { it.value.first < minVal }
-            trim(maxSize, minVal + 1) // recursion ftw
-        }
-        return this
-    }
+package org.ivdnt.galahad.evaluation.distribution
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import org.ivdnt.galahad.data.layer.Term
+
+/**
+ * Generic class for frequency distributions of terms in a corpus or document.
+ * The idea is to sum up the distribution as we go through the terms one by one using [add].
+ */
+open class Distribution {
+
+    /**
+     * (lem, pos) -> (count, literal[])
+     */
+    @JsonIgnore
+    val distributionMap: MutableMap<Pair<String, String>, Pair<Int, LiteralsEntry>> = HashMap()
+
+    var isTrimmed = false
+    var coveredChars = 0
+    var coveredAlphabeticChars = 0
+    var totalChars = 0
+    var totalAlphabeticChars = 0
+
+    /**
+     * Is serialized and send through API, so it is in fact used.
+     */
+    val distribution: Set<DistributionRow>
+        get() = distributionMap.entries.map {
+                DistributionRow(
+                    it.key.first,
+                    it.key.second,
+                    it.value.first,
+                    it.value.second
+                )
+            }.toSet()
+
+    fun add(term: Term) {
+        val literal: String = term.literals
+        coveredChars += literal.length
+        coveredAlphabeticChars += literal.count { char -> char.isLetter() }
+        add(
+            lemma = term.lemma ?: Term.NO_LEMMA,
+            pos = term.posHeadGroup ?: Term.NO_POS,
+            count = 1,
+            literals = LiteralsEntry(mapOf(term.literals to 1))
+        )
+    }
+
+    private fun add(lemma: String, pos: String, count: Int, literals: LiteralsEntry) {
+        distributionMap.merge(
+            Pair(lemma, pos), Pair(count, literals)
+        ) { p1, p2 -> Pair(p1.first + p2.first, p1.second.add(p2.second)) }
+    }
+
+    fun add(other: Distribution) {
+        coveredChars += other.coveredChars
+        coveredAlphabeticChars += other.coveredAlphabeticChars
+        totalChars += other.totalChars
+        totalAlphabeticChars += other.totalAlphabeticChars
+        other.distributionMap.forEach {
+            add(it.key.first, it.key.second, it.value.first, it.value.second)
+        }
+    }
+
+    fun toCSV(): String {
+        var csv: String = DistributionRow.getCsvHeader()
+        getSorted().forEach { csv += it.toCSVRecord() }
+        return csv
+    }
+
+    @JsonIgnore
+    fun getSorted(): Set<DistributionRow> {
+        return distributionMap.toList()
+            // Sort on count (descending, hence minus), then on lemma.
+            // Note that 'it' is of type Pair< Pair<String, String> , Pair<Int, LiteralsEntry> >
+            .sortedWith(compareBy({ -it.second.first }, { it.first.first.lowercase() })).toMap()
+            // Map to serializable rows.
+            .entries.map { DistributionRow(it.key.first, it.key.second, it.value.first, it.value.second) }.toSet()
+    }
+
+    fun trim(maxSize: Int, minVal: Int = 2): Distribution {
+        if (distributionMap.size > maxSize) {
+            isTrimmed = true
+            distributionMap.entries.removeIf { it.value.first < minVal }
+            trim(maxSize, minVal + 1) // recursion ftw
+        }
+        return this
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/Metric.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/Metric.kt
index 1435ae5..dadde11 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/Metric.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/Metric.kt
@@ -1,164 +1,166 @@
-package org.ivdnt.galahad.evaluation.metrics
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import com.fasterxml.jackson.annotation.JsonProperty
-import org.ivdnt.galahad.evaluation.EvaluationEntry
-import org.ivdnt.galahad.port.csv.CSVFile
-import org.ivdnt.galahad.port.csv.CSVHeader
-import org.ivdnt.galahad.port.csv.CSVRecord
-import org.ivdnt.galahad.util.toFixed
-
-/**
- * The number of [Term]s between two [Layer]s that are equal, different or not present in the reference and hypothesis layers.
- * Based on lemma and part of speech.
- *
- * Note that normal [Metric]s always contain samples, which are cheap to get when we are calculating them anyway,
- * However for downstream applications like aggregates, we may want to omit the samples
- * Therefore we can use this utility class
- */
-
-data class ClassificationMetrics(
-    val precision: Float = 0f,
-    val recall: Float = 0f,
-    val f1: Float = 0f,
-    val accuracy: Float = 0f,
-) {
-    operator fun plus(other: ClassificationMetrics): ClassificationMetrics = ClassificationMetrics(
-        precision + other.precision,
-        recall + other.recall,
-        f1 + other.f1,
-        accuracy + other.accuracy,
-    )
-
-    operator fun div(divisor: Int): ClassificationMetrics = this * (1.0f / divisor)
-
-    operator fun times(factor: Float): ClassificationMetrics = ClassificationMetrics(
-        precision * factor,
-        recall * factor,
-        f1 * factor,
-        accuracy * factor,
-    )
-
-    companion object {
-        fun calculate(cls: ClassificationClasses): ClassificationMetrics {
-            return calculate(cls.flat)
-        }
-
-        fun calculate(cls: FlatClassificationClasses, micro: Boolean = false): ClassificationMetrics {
-            val tp = cls.truePositive.toFloat()
-            val fp = cls.falsePositive.toFloat()
-            val fn = cls.falseNegative.toFloat()
-            // When calculating micro-accuracy, tp and fp are the same, so don't double count.
-            val total = if(micro) tp + fp else cls.count.toFloat()
-            return calculate(tp, fp, fn, total)
-        }
-
-        private fun calculate(tp: Float, fp: Float, fn: Float, total: Float): ClassificationMetrics {
-            fun notNaN(value: Float): Float = if (value.isNaN()) 0.0f else value
-
-            val precision = notNaN(tp / (tp + fp))
-            val recall = notNaN(tp / (tp + fn))
-
-            return ClassificationMetrics(
-                precision = precision,
-                recall = recall,
-                f1 = notNaN(2.0f * (precision * recall) / (precision + recall)),
-                accuracy = notNaN(tp / (total)),
-            )
-        }
-    }
-}
-
-/** Use for micro-averaging */
-open class FlatClassificationClasses(
-    var truePositive: Int = 0,
-    var falsePositive: Int = 0,
-    var falseNegative: Int = 0,
-    // Similar to falseNegative.
-    var noMatch: Int = 0,
-    var count: Int = 0,
-) {
-    operator fun plus(other: FlatClassificationClasses) = FlatClassificationClasses(
-        truePositive + other.truePositive,
-        falsePositive + other.falsePositive,
-        falseNegative + other.falseNegative,
-        noMatch + other.noMatch,
-        count + other.count,
-    )
-}
-
-open class ClassificationClasses(
-    var truePositive: EvaluationEntry = EvaluationEntry(),
-    var falsePositive: EvaluationEntry = EvaluationEntry(),
-    var falseNegative: EvaluationEntry = EvaluationEntry(),
-    // Similar to falseNegative.
-    var noMatch: EvaluationEntry = EvaluationEntry(),
-    /** sample count without duplicates, for calculating accuracy. */
-    @JsonIgnore var count: Int = 1,
-) {
-    open val classCount: Int
-        get() = truePositive.count + falsePositive.count + falseNegative.count + noMatch.count
-
-    fun add(other: ClassificationClasses, truncate: Boolean = true): ClassificationClasses {
-        truePositive = EvaluationEntry.add(truePositive, other.truePositive, truncate)
-        falsePositive = EvaluationEntry.add(falsePositive, other.falsePositive, truncate)
-        falseNegative = EvaluationEntry.add(falseNegative, other.falseNegative, truncate)
-        noMatch = EvaluationEntry.add(noMatch, other.noMatch, truncate)
-        count += other.count
-        return this
-    }
-
-    @get:JsonIgnore
-    open val flat: FlatClassificationClasses
-        get() = FlatClassificationClasses(
-            truePositive = truePositive.count,
-            falsePositive = falsePositive.count,
-            falseNegative = falseNegative.count,
-            noMatch = noMatch.count,
-            count = count
-        )
-}
-
-data class Metric(
-    @JsonProperty("name") val name: String,
-    @JsonProperty("classes") var cls: ClassificationClasses = ClassificationClasses(),
-) {
-    @get:JsonProperty("metrics")
-    val clsMetrics
-        get() = ClassificationMetrics.calculate(cls)
-
-    fun add(other: Metric, truncate: Boolean): Metric {
-        cls.add(other.cls, truncate)
-        return this
-    }
-
-    fun toCSVRecord(): CSVRecord {
-        return CSVFile.toCSVRecord(listOf(
-            name,
-            clsMetrics.precision.toFixed(),
-            clsMetrics.recall.toFixed(),
-            clsMetrics.f1.toFixed(),
-            cls.classCount,
-            cls.truePositive.count,
-            cls.falsePositive.count,
-            cls.falseNegative.count,
-            cls.noMatch.count,
-        ))
-    }
-
-    companion object {
-        fun getCsvHeader(): CSVHeader {
-            return CSVFile.toCSVHeader(listOf(
-                "grouped by",
-                "precision",
-                "recall",
-                "f1",
-                "count",
-                "true positive count",
-                "false positive count",
-                "false negative count",
-                "no match count")
-            )
-        }
-    }
+package org.ivdnt.galahad.evaluation.metrics
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import com.fasterxml.jackson.annotation.JsonProperty
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.evaluation.EvaluationEntry
+import org.ivdnt.galahad.port.csv.CSVFile
+import org.ivdnt.galahad.port.csv.CSVHeader
+import org.ivdnt.galahad.port.csv.CSVRecord
+import org.ivdnt.galahad.util.toFixed
+
+/**
+ * The number of [Term]s between two [Layer]s that are equal, different or not present in the reference and hypothesis layers.
+ * Based on lemma and part of speech.
+ *
+ * Note that normal [Metric]s always contain samples, which are cheap to get when we are calculating them anyway,
+ * However for downstream applications like aggregates, we may want to omit the samples
+ * Therefore we can use this utility class
+ */
+
+data class ClassificationMetrics(
+    val precision: Float = 0f,
+    val recall: Float = 0f,
+    val f1: Float = 0f,
+    val accuracy: Float = 0f,
+) {
+    operator fun plus(other: ClassificationMetrics): ClassificationMetrics = ClassificationMetrics(
+        precision + other.precision,
+        recall + other.recall,
+        f1 + other.f1,
+        accuracy + other.accuracy,
+    )
+
+    operator fun div(divisor: Int): ClassificationMetrics = this * (1.0f / divisor)
+
+    operator fun times(factor: Float): ClassificationMetrics = ClassificationMetrics(
+        precision * factor,
+        recall * factor,
+        f1 * factor,
+        accuracy * factor,
+    )
+
+    companion object {
+        fun calculate(cls: ClassificationClasses): ClassificationMetrics {
+            return calculate(cls.flat)
+        }
+
+        fun calculate(cls: FlatClassificationClasses, micro: Boolean = false): ClassificationMetrics {
+            val tp = cls.truePositive.toFloat()
+            val fp = cls.falsePositive.toFloat()
+            val fn = cls.falseNegative.toFloat()
+            // When calculating micro-accuracy, tp and fp are the same, so don't double count.
+            val total = if(micro) tp + fp else cls.count.toFloat()
+            return calculate(tp, fp, fn, total)
+        }
+
+        private fun calculate(tp: Float, fp: Float, fn: Float, total: Float): ClassificationMetrics {
+            fun notNaN(value: Float): Float = if (value.isNaN()) 0.0f else value
+
+            val precision = notNaN(tp / (tp + fp))
+            val recall = notNaN(tp / (tp + fn))
+
+            return ClassificationMetrics(
+                precision = precision,
+                recall = recall,
+                f1 = notNaN(2.0f * (precision * recall) / (precision + recall)),
+                accuracy = notNaN(tp / (total)),
+            )
+        }
+    }
+}
+
+/** Use for micro-averaging */
+open class FlatClassificationClasses(
+    var truePositive: Int = 0,
+    var falsePositive: Int = 0,
+    var falseNegative: Int = 0,
+    // Similar to falseNegative.
+    var noMatch: Int = 0,
+    var count: Int = 0,
+) {
+    operator fun plus(other: FlatClassificationClasses) = FlatClassificationClasses(
+        truePositive + other.truePositive,
+        falsePositive + other.falsePositive,
+        falseNegative + other.falseNegative,
+        noMatch + other.noMatch,
+        count + other.count,
+    )
+}
+
+open class ClassificationClasses(
+    var truePositive: EvaluationEntry = EvaluationEntry(),
+    var falsePositive: EvaluationEntry = EvaluationEntry(),
+    var falseNegative: EvaluationEntry = EvaluationEntry(),
+    // Similar to falseNegative.
+    var noMatch: EvaluationEntry = EvaluationEntry(),
+    /** sample count without duplicates, for calculating accuracy. */
+    @JsonIgnore var count: Int = 1,
+) {
+    open val classCount: Int
+        get() = truePositive.count + falsePositive.count + falseNegative.count + noMatch.count
+
+    fun add(other: ClassificationClasses, truncate: Boolean = true): ClassificationClasses {
+        truePositive = EvaluationEntry.add(truePositive, other.truePositive, truncate)
+        falsePositive = EvaluationEntry.add(falsePositive, other.falsePositive, truncate)
+        falseNegative = EvaluationEntry.add(falseNegative, other.falseNegative, truncate)
+        noMatch = EvaluationEntry.add(noMatch, other.noMatch, truncate)
+        count += other.count
+        return this
+    }
+
+    @get:JsonIgnore
+    open val flat: FlatClassificationClasses
+        get() = FlatClassificationClasses(
+            truePositive = truePositive.count,
+            falsePositive = falsePositive.count,
+            falseNegative = falseNegative.count,
+            noMatch = noMatch.count,
+            count = count
+        )
+}
+
+data class Metric(
+    @JsonProperty("name") val name: String,
+    @JsonProperty("classes") var cls: ClassificationClasses = ClassificationClasses(),
+) {
+    @get:JsonProperty("metrics")
+    val clsMetrics
+        get() = ClassificationMetrics.calculate(cls)
+
+    fun add(other: Metric, truncate: Boolean): Metric {
+        cls.add(other.cls, truncate)
+        return this
+    }
+
+    fun toCSVRecord(): CSVRecord {
+        return CSVFile.toCSVRecord(listOf(
+            name,
+            clsMetrics.precision.toFixed(),
+            clsMetrics.recall.toFixed(),
+            clsMetrics.f1.toFixed(),
+            cls.classCount,
+            cls.truePositive.count,
+            cls.falsePositive.count,
+            cls.falseNegative.count,
+            cls.noMatch.count,
+        ))
+    }
+
+    companion object {
+        fun getCsvHeader(): CSVHeader {
+            return CSVFile.toCSVHeader(listOf(
+                "grouped by",
+                "precision",
+                "recall",
+                "f1",
+                "count",
+                "true positive count",
+                "false positive count",
+                "false negative count",
+                "no match count")
+            )
+        }
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/MetricsType.kt b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/MetricsType.kt
index 5633b8d..705d92c 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/MetricsType.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/evaluation/metrics/MetricsType.kt
@@ -1,177 +1,176 @@
-package org.ivdnt.galahad.evaluation.metrics
-
-import com.fasterxml.jackson.annotation.JsonIgnore
-import org.ivdnt.galahad.data.layer.Term
-import org.ivdnt.galahad.evaluation.CsvSampleExporter
-import org.ivdnt.galahad.evaluation.EvaluationEntry
-import org.ivdnt.galahad.evaluation.comparison.TermComparison
-import org.ivdnt.galahad.port.csv.CSVFile
-import org.ivdnt.galahad.port.csv.CSVHeader
-import org.ivdnt.galahad.util.toFixed
-
-typealias FlatMetricTypeAssay = Map<String, FlatMetricType>
-class FlatMetricType(
-    val micro: ClassificationMetrics = ClassificationMetrics(),
-    val macro: ClassificationMetrics = ClassificationMetrics(),
-)
-
-class MetricsType(
-    val setting: MetricsSettings,
-    @JsonIgnore var truncate: Boolean = true
-) : CsvSampleExporter {
-    @JsonIgnore
-    var map: MutableMap<String, Metric> = HashMap()
-
-    /** Metrics separated per POS. */
-    val grouped: Set<Metric>
-        get() {
-            return if (map.size > TRUNCATE) {
-                // sort mt.value.map on mt.value.map["someKey"].cls.classCount, take the first TRUNCATE elements, and then map
-                map.entries.asSequence()
-                        .sortedByDescending { it.value.cls.classCount }.take(TRUNCATE)
-                        .associateBy({ it.key }, { it.value }).values.toSet()
-            } else {
-                map.values.toSet()
-            }
-        }
-
-    val classes: ClassificationClasses
-        get() = map.values.map { it.cls }.toMutableList().apply{ this.add(0, ClassificationClasses(count=0)) }.reduce { a, b -> a.add(b, truncate) }.apply { falsePositive = EvaluationEntry() }
-
-    val macro: ClassificationMetrics
-        get() {
-            if (map.isEmpty()) {
-                return ClassificationMetrics()
-            }
-            return map.values.map { it.clsMetrics }.reduce { a, b -> a + b } / map.size
-        }
-
-    val micro: ClassificationMetrics
-        get() {
-            if (map.isEmpty()) {
-                return ClassificationMetrics()
-            }
-            return ClassificationMetrics.calculate(map.values.map { it.cls.flat }.reduce { a, b -> a + b }, micro = true)
-        }
-
-    fun toGlobalCsv(): String {
-        // Expensive calculations.
-        val microMetrics = micro
-        val macroMetrics = macro
-
-        return CSVFile.toCSVRecord(listOf(
-            setting.annotation,
-            setting.group,
-            macroMetrics.precision.toFixed(),
-            macroMetrics.recall.toFixed(),
-            macroMetrics.f1.toFixed(),
-            microMetrics.accuracy.toFixed(),
-            classes.classCount,
-            classes.truePositive.count,
-            classes.falseNegative.count,
-            classes.noMatch.count,
-        ))
-    }
-
-    fun toFlat(): FlatMetricType {
-        return FlatMetricType(micro, macro)
-    }
-
-    fun toGroupedCsv(): String {
-        var csv = Metric.getCsvHeader()
-        grouped.sortedBy { it.name }.forEach{ csv += it.toCSVRecord() }
-        return csv
-    }
-
-    // Cumulative addition functions.
-    private fun add(metric: Metric) {
-        map.merge(metric.name, metric) { m1, m2 -> m1.add(m2, truncate) }
-    }
-
-    fun add(other: MetricsType) {
-        other.map.values.toSet().forEach(this::add)
-    }
-
-    fun add(comp: TermComparison) {
-        if (!setting.filterBy(comp)) {
-            return
-        }
-
-        if (comp.hypoTerm == Term.EMPTY) {
-            add (
-                Metric(
-                    name = setting.groupBy(comp.refTerm),
-                    cls = ClassificationClasses(
-                        noMatch = EvaluationEntry(1, mutableListOf(comp)),
-                        count = 0
-                    )
-                )
-            )
-        }
-
-        // One of these two will be empty, we don't know which.
-        val (trues, falses) = truesFalses(comp, setting::termsEqual)
-        val cls = ClassificationClasses(
-            truePositive = trues,
-            falseNegative = falses,
-        )
-        add(
-            Metric(
-                name = setting.groupBy(comp.refTerm),
-                cls = cls
-            )
-        )
-        if (falses.count != 0) {
-            // This term is also be someone else's false positive, so switch around.
-            val cls2 = ClassificationClasses(
-                falsePositive = EvaluationEntry(count = falses.samples.size, falses.samples.toMutableList()),
-                count = if (setting.groupBy(comp.hypoTerm) == setting.groupBy(comp.refTerm)) 0 else 1
-            )
-            add(
-                Metric(
-                    name = setting.groupBy(comp.hypoTerm), // Terms are switched, so hypo.
-                    cls = cls2
-                )
-            )
-        }
-    }
-
-    private fun truesFalses(comp: TermComparison, cond: (TermComparison) -> Boolean): Pair<EvaluationEntry, EvaluationEntry> {
-        val trues = if (cond(comp)) {
-            EvaluationEntry(1, mutableListOf(comp))
-        } else {
-            EvaluationEntry()
-        }
-        val falses = if (!cond(comp)) {
-            EvaluationEntry(1, mutableListOf(comp))
-        } else {
-            EvaluationEntry()
-        }
-        return Pair(trues, falses)
-    }
-
-    fun samplesToCsv(group: String, classType: String): String {
-        return when (classType) {
-            "truePositive" -> samplesToCSV(map[group]?.cls?.truePositive?.samples)
-            "falsePositive" -> samplesToCSV(map[group]?.cls?.falsePositive?.samples)
-            "falseNegative" -> samplesToCSV(map[group]?.cls?.falseNegative?.samples)
-            "noMatch" -> samplesToCSV(map[group]?.cls?.noMatch?.samples)
-            else -> ""
-        }
-    }
-
-    fun samplesToCsv(classType: String): String {
-        return when (classType) {
-            "truePositive" -> samplesToCSV(classes.truePositive.samples)
-            "falseNegative" -> samplesToCSV(classes.falseNegative.samples)
-            "noMatch" -> samplesToCSV(classes.noMatch.samples)
-            else -> ""
-        }
-    }
-
-
-    override fun samplesToCSV(): String {
-        val cls = listOf(classes.falsePositive, classes.falseNegative, classes.truePositive)
-        return samplesToCSV(cls.flatMap { it.samples })
-    }
+package org.ivdnt.galahad.evaluation.metrics
+
+import com.fasterxml.jackson.annotation.JsonIgnore
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.evaluation.CsvSampleExporter
+import org.ivdnt.galahad.evaluation.EvaluationEntry
+import org.ivdnt.galahad.evaluation.comparison.TermComparison
+import org.ivdnt.galahad.port.csv.CSVFile
+import org.ivdnt.galahad.util.toFixed
+
+typealias FlatMetricTypeAssay = Map<String, FlatMetricType>
+class FlatMetricType(
+    val micro: ClassificationMetrics = ClassificationMetrics(),
+    val macro: ClassificationMetrics = ClassificationMetrics(),
+)
+
+class MetricsType(
+    val setting: MetricsSettings,
+    @JsonIgnore var truncate: Boolean = true
+) : CsvSampleExporter {
+    @JsonIgnore
+    var map: MutableMap<String, Metric> = HashMap()
+
+    /** Metrics separated per POS. */
+    val grouped: Set<Metric>
+        get() {
+            return if (map.size > TRUNCATE) {
+                // sort mt.value.map on mt.value.map["someKey"].cls.classCount, take the first TRUNCATE elements, and then map
+                map.entries.asSequence()
+                        .sortedByDescending { it.value.cls.classCount }.take(TRUNCATE)
+                        .associateBy({ it.key }, { it.value }).values.toSet()
+            } else {
+                map.values.toSet()
+            }
+        }
+
+    val classes: ClassificationClasses
+        get() = map.values.map { it.cls }.toMutableList().apply{ this.add(0, ClassificationClasses(count=0)) }.reduce { a, b -> a.add(b, truncate) }.apply { falsePositive = EvaluationEntry() }
+
+    val macro: ClassificationMetrics
+        get() {
+            if (map.isEmpty()) {
+                return ClassificationMetrics()
+            }
+            return map.values.map { it.clsMetrics }.reduce { a, b -> a + b } / map.size
+        }
+
+    val micro: ClassificationMetrics
+        get() {
+            if (map.isEmpty()) {
+                return ClassificationMetrics()
+            }
+            return ClassificationMetrics.calculate(map.values.map { it.cls.flat }.reduce { a, b -> a + b }, micro = true)
+        }
+
+    fun toGlobalCsv(): String {
+        // Expensive calculations.
+        val microMetrics = micro
+        val macroMetrics = macro
+
+        return CSVFile.toCSVRecord(listOf(
+            setting.annotation,
+            setting.group,
+            macroMetrics.precision.toFixed(),
+            macroMetrics.recall.toFixed(),
+            macroMetrics.f1.toFixed(),
+            microMetrics.accuracy.toFixed(),
+            classes.classCount,
+            classes.truePositive.count,
+            classes.falseNegative.count,
+            classes.noMatch.count,
+        ))
+    }
+
+    fun toFlat(): FlatMetricType {
+        return FlatMetricType(micro, macro)
+    }
+
+    fun toGroupedCsv(): String {
+        var csv = Metric.getCsvHeader()
+        grouped.sortedBy { it.name }.forEach{ csv += it.toCSVRecord() }
+        return csv
+    }
+
+    // Cumulative addition functions.
+    private fun add(metric: Metric) {
+        map.merge(metric.name, metric) { m1, m2 -> m1.add(m2, truncate) }
+    }
+
+    fun add(other: MetricsType) {
+        other.map.values.toSet().forEach(this::add)
+    }
+
+    fun add(comp: TermComparison) {
+        if (!setting.filterBy(comp)) {
+            return
+        }
+
+        if (comp.hypoTerm == Term.EMPTY) {
+            add (
+                Metric(
+                    name = setting.groupBy(comp.refTerm),
+                    cls = ClassificationClasses(
+                        noMatch = EvaluationEntry(1, mutableListOf(comp)),
+                        count = 0
+                    )
+                )
+            )
+        }
+
+        // One of these two will be empty, we don't know which.
+        val (trues, falses) = truesFalses(comp, setting::termsEqual)
+        val cls = ClassificationClasses(
+            truePositive = trues,
+            falseNegative = falses,
+        )
+        add(
+            Metric(
+                name = setting.groupBy(comp.refTerm),
+                cls = cls
+            )
+        )
+        if (falses.count != 0) {
+            // This term is also be someone else's false positive, so switch around.
+            val cls2 = ClassificationClasses(
+                falsePositive = EvaluationEntry(count = falses.samples.size, falses.samples.toMutableList()),
+                count = if (setting.groupBy(comp.hypoTerm) == setting.groupBy(comp.refTerm)) 0 else 1
+            )
+            add(
+                Metric(
+                    name = setting.groupBy(comp.hypoTerm), // Terms are switched, so hypo.
+                    cls = cls2
+                )
+            )
+        }
+    }
+
+    private fun truesFalses(comp: TermComparison, cond: (TermComparison) -> Boolean): Pair<EvaluationEntry, EvaluationEntry> {
+        val trues = if (cond(comp)) {
+            EvaluationEntry(1, mutableListOf(comp))
+        } else {
+            EvaluationEntry()
+        }
+        val falses = if (!cond(comp)) {
+            EvaluationEntry(1, mutableListOf(comp))
+        } else {
+            EvaluationEntry()
+        }
+        return Pair(trues, falses)
+    }
+
+    fun samplesToCsv(group: String, classType: String): String {
+        return when (classType) {
+            "truePositive" -> samplesToCSV(map[group]?.cls?.truePositive?.samples)
+            "falsePositive" -> samplesToCSV(map[group]?.cls?.falsePositive?.samples)
+            "falseNegative" -> samplesToCSV(map[group]?.cls?.falseNegative?.samples)
+            "noMatch" -> samplesToCSV(map[group]?.cls?.noMatch?.samples)
+            else -> ""
+        }
+    }
+
+    fun samplesToCsv(classType: String): String {
+        return when (classType) {
+            "truePositive" -> samplesToCSV(classes.truePositive.samples)
+            "falseNegative" -> samplesToCSV(classes.falseNegative.samples)
+            "noMatch" -> samplesToCSV(classes.noMatch.samples)
+            else -> ""
+        }
+    }
+
+
+    override fun samplesToCSV(): String {
+        val cls = listOf(classes.falsePositive, classes.falseNegative, classes.truePositive)
+        return samplesToCSV(cls.flatMap { it.samples })
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/jobs/DocumentJob.kt b/server/src/main/kotlin/org/ivdnt/galahad/jobs/DocumentJob.kt
index 046d4b8..6455ebd 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/jobs/DocumentJob.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/jobs/DocumentJob.kt
@@ -7,6 +7,7 @@ import org.ivdnt.galahad.FileBackedValue
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.layer.LayerPreview
 import org.ivdnt.galahad.data.layer.LayerSummary
+import org.ivdnt.galahad.jobs.DocumentJob.DocumentProcessingStatus
 import org.ivdnt.galahad.tagset.Tagset
 import java.io.File
 import java.util.*
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/jobs/InternalJobController.kt b/server/src/main/kotlin/org/ivdnt/galahad/jobs/InternalJobController.kt
index 0fc5429..ff93245 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/jobs/InternalJobController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/jobs/InternalJobController.kt
@@ -1,17 +1,19 @@
 package org.ivdnt.galahad.jobs
 
 import org.apache.logging.log4j.kotlin.Logging
-import org.ivdnt.galahad.app.*
+import org.ivdnt.galahad.app.Config
+import org.ivdnt.galahad.app.INTERNAL_JOBS_ERROR_URL
+import org.ivdnt.galahad.app.INTERNAL_JOBS_RESULT_URL
 import org.ivdnt.galahad.data.CorporaController
 import org.ivdnt.galahad.data.document.Document
-import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.port.SourceLayerableFile
+import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.port.InternalFile
+import org.ivdnt.galahad.port.SourceLayerableFile
 import org.ivdnt.galahad.port.tsv.TSVFile
 import org.ivdnt.galahad.taggers.Taggers
-import org.ivdnt.galahad.tagset.TagsetStore
 import org.ivdnt.galahad.tagset.Tagset
+import org.ivdnt.galahad.tagset.TagsetStore
 import org.springframework.web.bind.annotation.*
 import org.springframework.web.multipart.MultipartFile
 import java.io.File
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/jobs/JobsController.kt b/server/src/main/kotlin/org/ivdnt/galahad/jobs/JobsController.kt
index f0d46a2..4f56cb7 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/jobs/JobsController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/jobs/JobsController.kt
@@ -1,15 +1,13 @@
 package org.ivdnt.galahad.jobs
 
-import com.fasterxml.jackson.annotation.JsonProperty
+import jakarta.servlet.http.HttpServletRequest
+import jakarta.servlet.http.HttpServletResponse
 import org.ivdnt.galahad.app.JOBS_URL
 import org.ivdnt.galahad.app.JOB_URL
 import org.ivdnt.galahad.data.CorporaController
 import org.springframework.beans.factory.annotation.Autowired
 import org.springframework.web.bind.annotation.*
 import java.util.*
-import jakarta.servlet.http.HttpServletRequest
-import jakarta.servlet.http.HttpServletResponse
-import org.ivdnt.galahad.app.TAGGERS_URL
 
 @RestController
 class JobsController(
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/CmdiMetadata.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/CmdiMetadata.kt
index 120c843..13dc261 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/CmdiMetadata.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/CmdiMetadata.kt
@@ -1,60 +1,96 @@
-package org.ivdnt.galahad.port
-
-import org.ivdnt.galahad.data.corpus.CorpusMetadata
-import org.ivdnt.galahad.util.escapeXML
-import org.ivdnt.galahad.util.toNonEmptyString
-import java.io.File
-import java.text.SimpleDateFormat
-import java.util.*
-import kotlin.io.path.createTempDirectory
-
-class CmdiMetadata(transformMetadata: DocumentTransformMetadata) : LayerTransformer(transformMetadata) {
-    companion object {
-        val tmp_dir: File = createTempDirectory("cmdi").toFile()
-    }
-
-    val file: File
-
-    init {
-        var template = this::class.java.classLoader.getResource("CMDI-template.xml")!!.readText()
-        val corpusMetadata: CorpusMetadata = this.transformMetadata.corpus.metadata.expensiveGet()
-        val docTitle = document.getUploadedRawFile().nameWithoutExtension
-
-        // Current year, month and day, zero-padded
-        val year = SimpleDateFormat("yyyy").format(Date())
-        val month = SimpleDateFormat("MM").format(Date())
-        val day = SimpleDateFormat("dd").format(Date())
-        val date = "$year-$month-$day"
-
-        // Retrieve GaLAHaD version from the same version.yml used in the client about page.
-        val versionStream = this::class.java.classLoader.getResource("version.yml")!!.openStream()
-        val versionProperties = Properties()
-        versionProperties.load(versionStream)
-        val galahadVersion = versionProperties.getProperty("VERSION")
-
-        val replacements = mapOf<String, String>(
-            "CORPUS_NAME" to corpusMetadata.name,
-            "DATE" to date,
-            "YEAR" to year,
-            "MONTH" to month,
-            "DAY" to day,
-            "PID" to document.uuid.toString(),
-            "GALAHAD_VERSION" to galahadVersion,
-            "TITLE" to docTitle,
-            "SOURCE_NAME" to corpusMetadata.sourceName.toNonEmptyString("!No source name defined!"),
-            "SOURCE_URL" to corpusMetadata.sourceURL.toNonEmptyString("!No source URL defined!"),
-            "ERA_FROM" to corpusMetadata.eraFrom.toString(),
-            "ERA_TO" to corpusMetadata.eraTo.toString(),
-            "TAGSET" to tagger.tagset.toNonEmptyString("!No tagset defined!"),
-            "FORMAT" to document.format.identifier,
-            "TAGGER_NAME" to tagger.id,
-            "TAGGER_VERSION" to tagger.version,
-            "TAGGER_URL" to tagger.model.href,
-        )
-        for ((key, value) in replacements) {
-            template = template.replace(key, value.escapeXML())
-        }
-        file = tmp_dir.resolve("CMDI-$docTitle.xml")
-        file.writeText(template)
-    }
+package org.ivdnt.galahad.port
+
+import org.ivdnt.galahad.data.corpus.CorpusMetadata
+import org.ivdnt.galahad.tagset.TagsetStore
+import org.ivdnt.galahad.util.escapeXML
+import org.ivdnt.galahad.util.getXmlBuilder
+import org.ivdnt.galahad.util.toNonEmptyString
+import org.w3c.dom.Node
+import java.io.File
+import java.text.SimpleDateFormat
+import java.util.*
+import javax.xml.transform.Transformer
+import javax.xml.transform.TransformerFactory
+import javax.xml.transform.dom.DOMSource
+import javax.xml.transform.stream.StreamResult
+import javax.xml.xpath.XPathConstants
+import javax.xml.xpath.XPathFactory
+import kotlin.io.path.createTempDirectory
+
+/** Constructs a CMDI file for exported documents. */
+class CmdiMetadata(transformMetadata: DocumentTransformMetadata) : LayerTransformer(transformMetadata) {
+
+    companion object {
+        private val tmp_dir: File = createTempDirectory("cmdi").toFile()
+    }
+
+    /** We need tagsets to go from tagger.tagset to tagset.fullName */
+    private val tagsets = TagsetStore()
+
+    // Some vals for repeated access.
+    private val docTitle = document.getUploadedRawFile().nameWithoutExtension
+    private val corpusMetadata: CorpusMetadata = transformMetadata.corpus.metadata.expensiveGet()
+    private val format = transformMetadata.targetFormat.identifier
+
+    /** After initialization this file will contain the CMDI */
+    val file: File
+
+    init {
+        // Load CMDI template
+        val cmdiTemplate = this::class.java.classLoader.getResourceAsStream("CMDI-template.xml")
+        val xmlDoc = getXmlBuilder().parse(cmdiTemplate)
+
+        val replacements: Map<List<String>, String> = getReplacements()
+        // Replace them
+        for ((keys, value) in replacements) {
+            val xpath = XPathFactory.newInstance().newXPath()
+            for (key in keys) {
+                val expr = xpath.compile("CMD//$key")
+                val node = expr.evaluate(xmlDoc, XPathConstants.NODE) as Node
+                node.textContent = value.escapeXML()
+            }
+        }
+        // Write to disk
+        file = tmp_dir.resolve("CMDI-$docTitle.xml")
+        val tf: Transformer = TransformerFactory.newInstance().newTransformer()
+        tf.transform(DOMSource(xmlDoc), StreamResult(file.outputStream()))
+    }
+
+    private fun getReplacements(): Map<List<String>, String> {
+        // Current year, month and day, zero-padded
+        val now = Date()
+        val year = SimpleDateFormat("yyyy").format(now)
+        val month = SimpleDateFormat("MM").format(now)
+        val day = SimpleDateFormat("dd").format(now)
+        val date = "$year-$month-$day"
+
+        // Retrieve GaLAHaD version from the same version.yml used in the client about page.
+        val versionStream = this::class.java.classLoader.getResource("version.yml")!!.openStream()
+        val versionProperties = Properties()
+        versionProperties.load(versionStream)
+        val galahadVersion = versionProperties.getProperty("GITHUB_REF_NAME")
+
+        // Define replacements
+        return mapOf(
+            listOf("MdCollectionDisplayName", "corpusName") to corpusMetadata.name,
+            listOf("MdCreationDate") to date,
+            listOf("Annotation//yearFrom", "Annotation//yearTo") to year,
+            listOf("Annotation//monthFrom", "Annotation//monthTo") to "--$month",
+            listOf("Annotation//dayFrom", "Annotation//dayTo") to "---$day",
+            listOf("ResourceRef") to "https://resolver.ivdnt.org/${document.uuid}",
+            listOf("GaLAHaDPersistentIdentifier") to "${document.uuid}_tei",
+            listOf("conversionDescription") to "exported to $format by GaLAHaD",
+            listOf("Conversion_GaLAHaD//toolVersion") to galahadVersion,
+            listOf("sourceID") to docTitle,
+            listOf("sourceCollection") to corpusMetadata.sourceName.toNonEmptyString("!No source name defined!"),
+            listOf("sourceCollectionURI") to corpusMetadata.sourceURL.toNonEmptyString("!No source URL defined!"),
+            listOf("Source_GaLAHaD//yearFrom") to corpusMetadata.eraFrom.toString(),
+            listOf("Source_GaLAHaD//yearTo") to corpusMetadata.eraTo.toString(),
+            listOf("annotationSet") to tagsets.getOrNull(tagger.tagset)?.longName.toNonEmptyString("!No tagset defined!"),
+            listOf("annotationFormat") to format,
+            listOf("Annotation//toolName") to tagger.id,
+            listOf("Annotation//toolVersion") to tagger.version,
+            listOf("Annotation//toolURI") to tagger.model.href,
+        )
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/TransformMetadata.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/TransformMetadata.kt
index 4c4951a..c719716 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/TransformMetadata.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/TransformMetadata.kt
@@ -1,47 +1,51 @@
-package org.ivdnt.galahad.port
-
-import org.ivdnt.galahad.data.corpus.Corpus
-import org.ivdnt.galahad.app.User
-import org.ivdnt.galahad.data.document.Document
-import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.data.layer.Term
-import org.ivdnt.galahad.jobs.Job
-
-open class CorpusTransformMetadata(
-    val corpus: Corpus,
-    val job: Job,
-    val user: User
-) {
-
-    fun documentMetadata( document: String ): DocumentTransformMetadata {
-        return DocumentTransformMetadata(
-            corpus = corpus,
-            job = job,
-            document = corpus.documents.readOrThrow( document ),
-            user = user
-        )
-    }
-}
-
-class DocumentTransformMetadata(
-    val corpus: Corpus,
-    val job: Job,
-    val document: Document,
-    val user: User
-) {
-
-    val layer: Layer = job.document(document.name).result
-
-    val plainText: String
-        get() = document.plaintext
-
-    fun convertLayerToPosHead() {
-        for (i in layer.terms.indices) {
-            val t = layer.terms[i]
-            layer.terms[i] = Term(
-                lemma = t.lemma,
-                pos = t.posHeadGroup,
-                targets = t.targets)
-        }
-    }
-}
+package org.ivdnt.galahad.port
+
+import org.ivdnt.galahad.app.User
+import org.ivdnt.galahad.data.corpus.Corpus
+import org.ivdnt.galahad.data.document.Document
+import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.jobs.Job
+
+open class CorpusTransformMetadata(
+    val corpus: Corpus,
+    val job: Job,
+    val user: User,
+    val targetFormat: DocumentFormat,
+) {
+
+    fun documentMetadata( document: String ): DocumentTransformMetadata {
+        return DocumentTransformMetadata(
+            corpus = corpus,
+            job = job,
+            document = corpus.documents.readOrThrow( document ),
+            user = user,
+            targetFormat = targetFormat,
+        )
+    }
+}
+
+class DocumentTransformMetadata(
+    val corpus: Corpus,
+    val job: Job,
+    val document: Document,
+    val user: User,
+    val targetFormat: DocumentFormat
+) {
+
+    val layer: Layer = job.document(document.name).result
+
+    val plainText: String
+        get() = document.plaintext
+
+    fun convertLayerToPosHead() {
+        for (i in layer.terms.indices) {
+            val t = layer.terms[i]
+            layer.terms[i] = Term(
+                lemma = t.lemma,
+                pos = t.posHeadGroup,
+                targets = t.targets)
+        }
+    }
+}
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/csv/CSVFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/csv/CSVFile.kt
index 6615ea3..7a68c0a 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/csv/CSVFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/csv/CSVFile.kt
@@ -10,7 +10,7 @@ class CSVFile(
 ) : File(path.toURI()) {
 
     init {
-        this.appendText(getExcelCompatibilityHeader())
+        this.appendText(excelCompatibilityHeader)
     }
 
     /** Append Excel compatible text. */
@@ -33,12 +33,13 @@ class CSVFile(
         // This is sad and I don't know of a clean solution short of having a configuration option 'excel-compatibility'
         // We will just except this loss and accommodate Excel/Windows user who might be less technically skilled
         // and hope that the Linux crowd will be able to handle this slightly weird header themselves
-        fun getExcelCompatibilityHeader(): String {
-            // Force Excel to read the csv as UTF16LE. Needed to render e.g. 'ü'.
-            // https://en.wikipedia.org/wiki/Byte_order_mark
-            val bom = '\uFEFF'
-            return "${bom}sep=,\n"
-        }
+        val excelCompatibilityHeader: String
+            get() {
+                // Force Excel to read the csv as UTF16LE. Needed to render e.g. 'ü'.
+                // https://en.wikipedia.org/wiki/Byte_order_mark
+                val bom = '\uFEFF'
+                return "${bom}sep=,\n"
+            }
 
         fun toCSVHeader(headers: List<String>): CSVHeader {
             // This is just an alias
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/FoliaReader.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/FoliaReader.kt
index 4b9a4d7..eda7688 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/FoliaReader.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/FoliaReader.kt
@@ -1,9 +1,9 @@
 package org.ivdnt.galahad.port.folia
 
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.layer.Term
 import org.ivdnt.galahad.data.layer.WordForm
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.port.folia.export.deepcopy
 import org.ivdnt.galahad.port.xml.reparseText
 import org.ivdnt.galahad.port.xml.tagName
@@ -18,8 +18,7 @@ class FoliaReader(
     val file: File,
     val nodeHandler: (node: Node, offset: Int, document: Document) -> Unit,
 ) {
-
-    val xmlDoc = getXmlBuilder().parse(file)
+    val xmlDoc: Document = getXmlBuilder().parse(file)
     val plainTextBuilder: StringBuilder = StringBuilder()
     val sourceLayer = Layer(SOURCE_LAYER_NAME)
 
@@ -83,9 +82,9 @@ class FoliaReader(
                 }
 
                 "s" -> {
-                    addNonFloatingString("\n")
+                    nonFloatingNL()
                     recurse()
-                    addNonFloatingString("\n")
+                    nonFloatingNL()
                     previousWasW = false
                 }
                 "p" -> {
@@ -103,18 +102,15 @@ class FoliaReader(
     }
 
     /** Adds a newline if the last character exists and is a non newline.*/
-    private fun addNonFloatingString(str: String) {
-        if (plainTextBuilder.isNotEmpty() && !plainTextBuilder.endsWith(str)) {
-            plainTextBuilder.append(str)
+    private fun nonFloatingNL() {
+        if (plainTextBuilder.isNotEmpty() && !plainTextBuilder.endsWith("\n")) {
+            plainTextBuilder.append("\n")
         }
     }
 
     private fun nonFloatingDoubleNL() {
-        if (plainTextBuilder.isEmpty())
-            return
-        else if (!plainTextBuilder.endsWith("\n")) {
-            plainTextBuilder.append("\n\n")
-        } else if (!plainTextBuilder.endsWith("\n\n")) {
+        nonFloatingNL()
+        if (plainTextBuilder.isNotEmpty() && !plainTextBuilder.endsWith("\n\n")) {
             plainTextBuilder.append("\n")
         }
     }
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/FoliaTextMerger.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/FoliaTextMerger.kt
index 7e8da15..9125da0 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/FoliaTextMerger.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/FoliaTextMerger.kt
@@ -1,151 +1,150 @@
-package org.ivdnt.galahad.port.folia.export
-
-import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.data.layer.Term
-import org.ivdnt.galahad.data.layer.WordForm
-import org.ivdnt.galahad.port.tei.export.TEITextMerger
-import org.ivdnt.galahad.util.insertFirst
-import org.ivdnt.galahad.port.xml.reparseText
-import org.ivdnt.galahad.port.xml.tagName
-import org.ivdnt.galahad.util.childOrNull
-import org.ivdnt.galahad.util.insertAfter
-import org.w3c.dom.Document
-import org.w3c.dom.Element
-import org.w3c.dom.Node
-import org.w3c.dom.NodeList
-
-fun NodeList.deepcopy(): ArrayList<Node> {
-    val copy = ArrayList<Node>()
-    for (i in 0 until this.length) {
-        copy.add(this.item(i))
-    }
-    return copy
-}
-
-class FoliaTextMerger(
-    node: Node, offset: Int, document: Document, wordFormIter: ListIterator<WordForm>,
-    deleteList: ArrayList<Node>, layer: Layer
-) : TEITextMerger(node, offset, document, wordFormIter, deleteList, layer, DocumentFormat.Folia) {
-
-    override fun merge() {
-        if (node.tagName() == "t" || node.tagName()?.startsWith("t-") == true) {
-            // We are going to add nodes when exporting, but we don't want to iterate over them
-            // So better store references to the current nodes and use them
-            val parent = node
-            val oldChildNodes = node.childNodes.deepcopy()
-            var endsWithSpace = true
-            for (child in oldChildNodes) {
-                node = child
-                val text = reparseText(child.textContent)
-
-                if (node.nodeType == Node.TEXT_NODE) {
-                    child.textContent = text
-                }
-                // TODO reparseText overwrites embedded t-styles
-
-                // never set the offset of more than one space.
-                if (endsWithSpace && text.startsWith(" ")) {
-                    offset -= 1
-                }
-                merge()
-
-                // Keep track of the ending space
-                if (text.isNotEmpty() && text.endsWith(" "))
-                    endsWithSpace = true
-                else if (text.isNotEmpty())
-                    endsWithSpace = false
-
-                if (child.nodeType == Node.TEXT_NODE)
-                    offset += text.length
-            }
-            // Remove parent and transfer children.
-            if(markForDeletion(parent)) {
-                var last = parent
-                for (i in parent.childNodes.length - 1 downTo 0) {
-                    val c = parent.childNodes.item(i)
-                    parent.parentNode.insertBefore(c, last)
-                    last = c
-                }
-            }
-
-        } else {
-            super.merge()
-        }
-    }
-
-    override fun createWTag(wf: WordForm): Element {
-        val wTag = node.parentNode.cloneNode(false)
-        return wTag as Element
-    }
-
-    override fun addWordForm(previousEndOffset: Int, wf: WordForm) {
-        super.addWordForm(previousEndOffset, wf)
-        // For Folia, newWTag is actually a <t> or <t-...> tag.
-        var tTag: Node = newWTag!!
-        // Make sure tTag points to a <t>. For e.g. a <t-style>, grab the first <t> parent.
-        var parent =
-            if (tTag.parentNode.tagName() == "t") tTag.parentNode
-            else tTag.parentNode.parentNode // First iteration looks at grandparent, because t-style copied itself.
-        while (tTag.tagName() != "t") {
-            val clone = parent.cloneNode(false)
-            tTag.parentNode.replaceChild(clone,tTag)
-            clone.insertFirst(tTag)
-            // Ready for next iter.
-            parent = parent.parentNode
-            tTag = clone
-        }
-        // Create the <w> which will contain the <t>
-        val wTag =  document.createElement("w")
-        val term = layer.termForWordForm(wf)
-        wTag.addTerm(term)
-        // Contain it.
-        tTag.parentNode.replaceChild(wTag,tTag)
-        wTag.insertFirst(tTag)
-    }
-
-    override fun handleElementNode() {
-        val element = node as Element
-        if (element.tagName != "w") return
-
-        val wordFormToAdd = getWordFormForOffsetOrNull() ?: return
-        val term = layer.termForWordForm(wordFormToAdd)
-        element.addTerm(term)
-    }
-
-    override fun moveWTagUp(wTag: Element): Element {
-        wTag.parentNode.parentNode.insertAfter(wTag, wTag.parentNode)
-        val clone = wTag.parentNode.cloneNode(false)
-        wTag.parentNode.replaceChild(clone, wTag)
-        clone.appendChild(wTag)
-        newWTag = clone as Element
-        return clone
-    }
-
-    private fun Element.addTerm(term: Term) {
-        this.addTermFeature("lemma", term.lemma.toString())
-        this.addTermFeature("pos", term.pos.toString(), term.posHeadGroup)
-    }
-
-    private fun Element.addTermFeature(name: String, value: String, head: String? = null) {
-        /* If at some point we want to remove existing annotations layers (pos & lemma) in folia <w> tags
-         * uncomment this. For now, multiple annotation layers are okay in the export.
-        // Find the child elements of [name] and delete them
-        val children = this.childNodes.deepcopy()
-        for (child in children) {
-            if (child.tagName() == name) {
-                this.removeChild(child)
-            }
-        }*/
-
-        // Create a new child element of [name]
-        val child = this.ownerDocument.createElement(name)
-        child.setAttribute("class", value)
-        // For PoS
-        if (head != null) child.setAttribute("head", head)
-        // Folia metadata.
-        child.setAttribute("processor", layer.name)
-        child.setAttribute("set", layer.name)
-        this.appendChild(child)
-    }
+package org.ivdnt.galahad.port.folia.export
+
+import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.data.layer.WordForm
+import org.ivdnt.galahad.port.tei.export.TEITextMerger
+import org.ivdnt.galahad.port.xml.reparseText
+import org.ivdnt.galahad.port.xml.tagName
+import org.ivdnt.galahad.util.insertAfter
+import org.ivdnt.galahad.util.insertFirst
+import org.w3c.dom.Document
+import org.w3c.dom.Element
+import org.w3c.dom.Node
+import org.w3c.dom.NodeList
+
+fun NodeList.deepcopy(): ArrayList<Node> {
+    val copy = ArrayList<Node>()
+    for (i in 0 until this.length) {
+        copy.add(this.item(i))
+    }
+    return copy
+}
+
+class FoliaTextMerger(
+    node: Node, offset: Int, document: Document, wordFormIter: ListIterator<WordForm>,
+    deleteList: ArrayList<Node>, layer: Layer
+) : TEITextMerger(node, offset, document, wordFormIter, deleteList, layer, DocumentFormat.Folia) {
+
+    override fun merge() {
+        if (node.tagName() == "t" || node.tagName()?.startsWith("t-") == true) {
+            // We are going to add nodes when exporting, but we don't want to iterate over them
+            // So better store references to the current nodes and use them
+            val parent = node
+            val oldChildNodes = node.childNodes.deepcopy()
+            var endsWithSpace = true
+            for (child in oldChildNodes) {
+                node = child
+                val text = reparseText(child.textContent)
+
+                if (node.nodeType == Node.TEXT_NODE) {
+                    child.textContent = text
+                }
+                // TODO reparseText overwrites embedded t-styles
+
+                // never set the offset of more than one space.
+                if (endsWithSpace && text.startsWith(" ")) {
+                    offset -= 1
+                }
+                merge()
+
+                // Keep track of the ending space
+                if (text.isNotEmpty() && text.endsWith(" "))
+                    endsWithSpace = true
+                else if (text.isNotEmpty())
+                    endsWithSpace = false
+
+                if (child.nodeType == Node.TEXT_NODE)
+                    offset += text.length
+            }
+            // Remove parent and transfer children.
+            if(markForDeletion(parent)) {
+                var last = parent
+                for (i in parent.childNodes.length - 1 downTo 0) {
+                    val c = parent.childNodes.item(i)
+                    parent.parentNode.insertBefore(c, last)
+                    last = c
+                }
+            }
+
+        } else {
+            super.merge()
+        }
+    }
+
+    override fun createWTag(wf: WordForm): Element {
+        val wTag = node.parentNode.cloneNode(false)
+        return wTag as Element
+    }
+
+    override fun addWordForm(previousEndOffset: Int, wf: WordForm) {
+        super.addWordForm(previousEndOffset, wf)
+        // For Folia, newWTag is actually a <t> or <t-...> tag.
+        var tTag: Node = newWTag!!
+        // Make sure tTag points to a <t>. For e.g. a <t-style>, grab the first <t> parent.
+        var parent =
+            if (tTag.parentNode.tagName() == "t") tTag.parentNode
+            else tTag.parentNode.parentNode // First iteration looks at grandparent, because t-style copied itself.
+        while (tTag.tagName() != "t") {
+            val clone = parent.cloneNode(false)
+            tTag.parentNode.replaceChild(clone,tTag)
+            clone.insertFirst(tTag)
+            // Ready for next iter.
+            parent = parent.parentNode
+            tTag = clone
+        }
+        // Create the <w> which will contain the <t>
+        val wTag =  document.createElement("w")
+        val term = layer.termForWordForm(wf)
+        wTag.addTerm(term)
+        // Contain it.
+        tTag.parentNode.replaceChild(wTag,tTag)
+        wTag.insertFirst(tTag)
+    }
+
+    override fun handleElementNode() {
+        val element = node as Element
+        if (element.tagName != "w") return
+
+        val wordFormToAdd = getWordFormForOffsetOrNull() ?: return
+        val term = layer.termForWordForm(wordFormToAdd)
+        element.addTerm(term)
+    }
+
+    override fun moveWTagUp(wTag: Element): Element {
+        wTag.parentNode.parentNode.insertAfter(wTag, wTag.parentNode)
+        val clone = wTag.parentNode.cloneNode(false)
+        wTag.parentNode.replaceChild(clone, wTag)
+        clone.appendChild(wTag)
+        newWTag = clone as Element
+        return clone
+    }
+
+    private fun Element.addTerm(term: Term) {
+        this.addTermFeature("lemma", term.lemmaOrEmpty)
+        this.addTermFeature("pos", term.posOrEmpty, term.posHeadGroupOrEmpty)
+    }
+
+    private fun Element.addTermFeature(name: String, value: String, head: String? = null) {
+        /* If at some point we want to remove existing annotations layers (pos & lemma) in folia <w> tags
+         * uncomment this. For now, multiple annotation layers are okay in the export.
+        // Find the child elements of [name] and delete them
+        val children = this.childNodes.deepcopy()
+        for (child in children) {
+            if (child.tagName() == name) {
+                this.removeChild(child)
+            }
+        }*/
+
+        // Create a new child element of [name]
+        val child = this.ownerDocument.createElement(name)
+        child.setAttribute("class", value)
+        // For PoS
+        if (head != null) child.setAttribute("head", head)
+        // Folia metadata.
+        child.setAttribute("processor", layer.name)
+        child.setAttribute("set", layer.name)
+        this.appendChild(child)
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/LayerToFoliaConverter.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/LayerToFoliaConverter.kt
index 5eecf59..777de28 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/LayerToFoliaConverter.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/folia/export/LayerToFoliaConverter.kt
@@ -1,82 +1,82 @@
-package org.ivdnt.galahad.port.folia.export
-
-import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.layer.Term
-import org.ivdnt.galahad.port.DocumentTransformMetadata
-import org.ivdnt.galahad.port.LayerConverter
-import org.ivdnt.galahad.port.LayerTransformer
-import org.ivdnt.galahad.util.XMLWriter
-import org.ivdnt.galahad.util.escapeXML
-import org.ivdnt.galahad.util.toValidXmlId
-import java.io.OutputStream
-
-class LayerToFoliaConverter (
-    transformMetadata: DocumentTransformMetadata,
-) : LayerConverter, LayerTransformer( transformMetadata ) {
-
-    override val format: DocumentFormat
-        get() = DocumentFormat.Folia
-
-    val id: String
-        get() = document.getUploadedRawFile().nameWithoutExtension.toValidXmlId()
-
-    override fun convert(outputStream: OutputStream) {
-        val taggerName = tagger.id
-        val writer = XMLWriter(outputStream)
-        // XML Header
-        writer.writeLineRaw("<?xml version=\"1.0\" encoding=\"UTF-8\"?>")
-        writer.openTag("<FoLiA xmlns:xlink=\"http://www.w3.org/1999/xlink\" xmlns=\"http://ilk.uvt.nl/folia\" xml:id=\"$id\" generator=\"galahad.ivdnt.org\" version=\"2.5.3\">")
-        // Metadata
-        writeMetadata(writer, taggerName)
-        // Textbody
-        writeTextBody(writer, taggerName)
-    }
-
-    private fun writeMetadata(writer: XMLWriter, taggerName: String) {
-        writer.openTag("<metadata type=\"native\">")
-        // Annotations
-        writer.openTag("<annotations>")
-        writer.writeLine("<text-annotation/>")
-        writer.writeLine("<paragraph-annotation/>")
-        writer.writeLine("<sentence-annotation/>")
-        writer.writeLine("<token-annotation/>")
-        for (annotation in setOf("lemma", "pos")) {
-            writer.openTag("<$annotation-annotation set=\"${taggerName}\">")
-            writer.writeLine("<annotator processor=\"${taggerName}\"/>")
-            writer.closeTag("</$annotation-annotation>")
-        }
-        writer.closeTag("</annotations>")
-        // Provenance
-        writer.openTag("<provenance>")
-        writer.writeLine(
-            "<processor host=\"galahad.ivdnt.org\" name=\"${taggerName}\" src=\"https://github.com/INL/taggers-dockerized\" type=\"auto\" user=\"${transformMetadata.user.id}\" xml:id=\"${taggerName}\"/>"
-        )
-        writer.closeTag("</provenance>")
-        writer.closeTag("</metadata>")
-    }
-
-    private fun writeTextBody(writer: XMLWriter, taggerName: String) {
-        writer.openTag("<text xml:id=\"$id.text\">")
-        writer.openTag("<p xml:id=\"$id.p1\">")
-        for ((index, term) in this.result.terms.withIndex()) {
-            // Single W
-            writeSingleW(writer, index, term, taggerName)
-        }
-        writer.closeTag("</p>")
-        writer.closeTag("</text>")
-        writer.closeTag("</FoLiA>")
-    }
-
-    private fun writeSingleW(
-        writer: XMLWriter, index: Int, term: Term,
-        taggerName: String,
-    ) {
-        writer.openTag("<w xml:id=\"$id.p1.w${index + 1}\">")
-        writer.writeLine("<t>${term.targets[0].literal.escapeXML()}</t>")
-        writer.writeLine("<lemma class=\"${term.lemma?.escapeXML()}\" processor=\"$taggerName\" set=\"$taggerName\"/>")
-        writer.writeLine(
-            "<pos class=\"${term.pos?.escapeXML()}\" head=\"${term.posHeadGroup?.escapeXML()}\" processor=\"$taggerName\" set=\"$taggerName\"/>"
-        )
-        writer.closeTag("</w>")
-    }
+package org.ivdnt.galahad.port.folia.export
+
+import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.layer.Term
+import org.ivdnt.galahad.port.DocumentTransformMetadata
+import org.ivdnt.galahad.port.LayerConverter
+import org.ivdnt.galahad.port.LayerTransformer
+import org.ivdnt.galahad.util.XMLWriter
+import org.ivdnt.galahad.util.escapeXML
+import org.ivdnt.galahad.util.toValidXmlId
+import java.io.OutputStream
+
+class LayerToFoliaConverter (
+    transformMetadata: DocumentTransformMetadata,
+) : LayerConverter, LayerTransformer( transformMetadata ) {
+
+    override val format: DocumentFormat
+        get() = DocumentFormat.Folia
+
+    val id: String
+        get() = document.getUploadedRawFile().nameWithoutExtension.toValidXmlId()
+
+    override fun convert(outputStream: OutputStream) {
+        val taggerName = tagger.id
+        val writer = XMLWriter(outputStream)
+        // XML Header
+        writer.writeLineRaw("<?xml version=\"1.0\" encoding=\"UTF-8\"?>")
+        writer.openTag("<FoLiA xmlns:xlink=\"http://www.w3.org/1999/xlink\" xmlns=\"http://ilk.uvt.nl/folia\" xml:id=\"$id\" generator=\"galahad.ivdnt.org\" version=\"2.5.3\">")
+        // Metadata
+        writeMetadata(writer, taggerName)
+        // Textbody
+        writeTextBody(writer, taggerName)
+    }
+
+    private fun writeMetadata(writer: XMLWriter, taggerName: String) {
+        writer.openTag("<metadata type=\"native\">")
+        // Annotations
+        writer.openTag("<annotations>")
+        writer.writeLine("<text-annotation/>")
+        writer.writeLine("<paragraph-annotation/>")
+        writer.writeLine("<sentence-annotation/>")
+        writer.writeLine("<token-annotation/>")
+        for (annotation in setOf("lemma", "pos")) {
+            writer.openTag("<$annotation-annotation set=\"${taggerName}\">")
+            writer.writeLine("<annotator processor=\"${taggerName}\"/>")
+            writer.closeTag("</$annotation-annotation>")
+        }
+        writer.closeTag("</annotations>")
+        // Provenance
+        writer.openTag("<provenance>")
+        writer.writeLine(
+            "<processor host=\"galahad.ivdnt.org\" name=\"${taggerName}\" src=\"https://github.com/INL/taggers-dockerized\" type=\"auto\" user=\"${transformMetadata.user.id}\" xml:id=\"${taggerName}\"/>"
+        )
+        writer.closeTag("</provenance>")
+        writer.closeTag("</metadata>")
+    }
+
+    private fun writeTextBody(writer: XMLWriter, taggerName: String) {
+        writer.openTag("<text xml:id=\"$id.text\">")
+        writer.openTag("<p xml:id=\"$id.p1\">")
+        for ((index, term) in this.result.terms.withIndex()) {
+            // Single W
+            writeSingleW(writer, index, term, taggerName)
+        }
+        writer.closeTag("</p>")
+        writer.closeTag("</text>")
+        writer.closeTag("</FoLiA>")
+    }
+
+    private fun writeSingleW(
+        writer: XMLWriter, index: Int, term: Term,
+        taggerName: String,
+    ) {
+        writer.openTag("<w xml:id=\"$id.p1.w${index + 1}\">")
+        writer.writeLine("<t>${term.targets[0].literal.escapeXML()}</t>")
+        writer.writeLine("<lemma class=\"${term.lemmaOrEmpty.escapeXML()}\" processor=\"$taggerName\" set=\"$taggerName\"/>")
+        writer.writeLine(
+            "<pos class=\"${term.posOrEmpty.escapeXML()}\" head=\"${term.posHeadGroupOrEmpty.escapeXML()}\" processor=\"$taggerName\" set=\"$taggerName\"/>"
+        )
+        writer.closeTag("</w>")
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/naf/NAFFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/naf/NAFFile.kt
index 37063bf..f287c73 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/naf/NAFFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/naf/NAFFile.kt
@@ -1,10 +1,10 @@
 package org.ivdnt.galahad.port.naf
 
 import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.layer.Term
 import org.ivdnt.galahad.data.layer.WordForm
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.port.DocumentTransformMetadata
 import org.ivdnt.galahad.port.PlainTextableFile
 import org.ivdnt.galahad.port.SourceLayerableFile
@@ -26,11 +26,11 @@ class NAFFile (
 
     val xmlDoc: Document = getXmlBuilder().parse(file)
 
-    val xPathfactory = XPathFactory.newInstance()
-    val xpath = xPathfactory.newXPath()
-    val expr = xpath.compile("/NAF/raw")
-    val wfExpr = xpath.compile("/NAF/text/wf")
-    val termExpr = xpath.compile("/NAF/terms/term")
+    private val xPathfactory = XPathFactory.newInstance()
+    private val xpath = xPathfactory.newXPath()
+    private val expr = xpath.compile("/NAF/raw")
+    private val wfExpr = xpath.compile("/NAF/text/wf")
+    private val termExpr = xpath.compile("/NAF/terms/term")
 
     override fun plainTextReader(): StringReader {
             return (expr.evaluate(xmlDoc, XPathConstants.NODE ) as Node).textContent.reader()
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/naf/export/LayerToNAFConverter.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/naf/export/LayerToNAFConverter.kt
index 0906d6b..4ff753b 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/naf/export/LayerToNAFConverter.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/naf/export/LayerToNAFConverter.kt
@@ -57,8 +57,8 @@ class LayerToNAFConverter (
                 val xterm = xmlDoc.createElement("term")
                 terms.appendChild(xterm)
                 xterm.setAttribute("id", "t$index")
-                xterm.setAttribute("lemma", term.lemma)
-                xterm.setAttribute("pos", term.pos)
+                xterm.setAttribute("lemma", term.lemmaOrEmpty)
+                xterm.setAttribute("pos", term.posOrEmpty)
 
                 val xspan = xmlDoc.createElement("span")
                 xterm.appendChild( xspan )
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/TEIFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/TEIFile.kt
index b601164..4edc2b3 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/TEIFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/TEIFile.kt
@@ -1,9 +1,9 @@
 package org.ivdnt.galahad.port.tei
 
 import org.ivdnt.galahad.app.executeAndLogTime
-import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.data.document.FormatInducer
+import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.port.DocumentTransformMetadata
 import org.ivdnt.galahad.port.tei.export.TEILayerMerger
 import org.ivdnt.galahad.port.xml.AnnotatedFile
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/LayerToTEIConverter.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/LayerToTEIConverter.kt
index 118b8f3..109c26d 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/LayerToTEIConverter.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/LayerToTEIConverter.kt
@@ -112,8 +112,8 @@ class LayerToTEIConverter(
             writer.writeRaw("<pc xml:id=\"${term.targets[0].id}\">${getLiteral()}</pc>")
         } else {
             // If it is not punctuation, safely assume it can be interpreted as <w>
-            val lemma = term.lemma?.escapeXML()
-            val pos = term.pos?.escapeXML()
+            val lemma = term.lemmaOrEmpty.escapeXML()
+            val pos = term.posOrEmpty.escapeXML()
             writer.writeRaw("<w lemma=\"$lemma\" pos=\"$pos\" xml:id=\"${term.targets[0].id}\">${getLiteral()}</w>")
         }
     }
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEIMetadata.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEIMetadata.kt
index 3d6af72..456c9cf 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEIMetadata.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEIMetadata.kt
@@ -5,15 +5,11 @@ import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.port.LayerTransformer
 import org.ivdnt.galahad.port.xml.XMLMetadata
 import org.ivdnt.galahad.util.childOrNull
-import org.ivdnt.galahad.util.getXmlBuilder
 import org.ivdnt.galahad.util.toNonEmptyString
 import org.w3c.dom.Document
 import org.w3c.dom.Element
 import org.w3c.dom.Node
-import org.xml.sax.InputSource
-import java.io.ByteArrayOutputStream
 import java.io.OutputStream
-import java.io.StringReader
 import java.util.*
 import javax.xml.transform.OutputKeys
 import javax.xml.transform.Transformer
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEITextMerger.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEITextMerger.kt
index 50cddb5..37f03e5 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEITextMerger.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tei/export/TEITextMerger.kt
@@ -1,17 +1,16 @@
 package org.ivdnt.galahad.port.tei.export
 
 import org.ivdnt.galahad.app.report.Report
+import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.data.layer.WordForm
+import org.ivdnt.galahad.evaluation.comparison.LayerComparison.Companion.truncatedPcMatch
 import org.ivdnt.galahad.port.folia.export.deepcopy
 import org.ivdnt.galahad.port.xml.getPlainTextContent
+import org.ivdnt.galahad.util.*
 import org.w3c.dom.Document
 import org.w3c.dom.Element
 import org.w3c.dom.Node
-import java.util.*
-import kotlin.collections.ArrayList
-import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.layer.WordForm
-import org.ivdnt.galahad.util.*
 
 fun HashSet<String>.contains(s: String?, ignoreCase: Boolean = false): Boolean {
     return any { it.equals(s, ignoreCase) }
@@ -234,11 +233,11 @@ open class TEITextMerger(
             n
         } else {
             val n = document.createElement("w")
-            n.setAttribute("lemma", termToAdd.lemma)
+            n.setAttribute("lemma", termToAdd.lemmaOrEmpty)
             n
         }
         // Both <w> and <pc> have a pos.
-        wTag.setAttribute(posType(), termToAdd.pos)
+        wTag.setAttribute(posType(), termToAdd.posOrEmpty)
         return wTag
     }
 
@@ -269,8 +268,9 @@ open class TEITextMerger(
             if (wordFormToAdd != null) {
                 // remove all whitespace within a <w>-tag (although this rarely occurs anyway).
                 val sourceLiteral = node.getPlainTextContent().replace(Regex("""\s"""), "")
-                if (wordFormToAdd.literal == sourceLiteral) {
-                    // This is a simple case since the tokenization matches
+                if (wordFormToAdd.literal == sourceLiteral // This is a simple case since the tokenization matches
+                    || truncatedPcMatch(sourceLiteral, wordFormToAdd.literal) // Also match with single punctuation (e.g. word. -> word)
+                ) {
                     mergeWTag(wordFormToAdd, element)
                 } else {
                     // Tokenization mismatch, report it
@@ -303,9 +303,9 @@ open class TEITextMerger(
         val termToAdd = layer.termForWordForm(wordFormToAdd)
         // <pc> tags do not have a lemma.
         if (element.tagName == "w") {
-            element.setAttribute("lemma", termToAdd.lemma)
+            element.setAttribute("lemma", termToAdd.lemmaOrEmpty)
         }
-        element.setAttribute(posType(), termToAdd.pos)
+        element.setAttribute(posType(), termToAdd.posOrEmpty)
         element.removeAttribute("type") // Update legacy formats to TEI p5
     }
 
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/TSVFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/TSVFile.kt
index 36d13fd..674f1ae 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/TSVFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/TSVFile.kt
@@ -1,9 +1,12 @@
 package org.ivdnt.galahad.port.tsv
 
 import org.ivdnt.galahad.data.document.DocumentFormat
-import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
-import org.ivdnt.galahad.port.*
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.port.DocumentTransformMetadata
+import org.ivdnt.galahad.port.InternalFile
+import org.ivdnt.galahad.port.PlainTextableFile
+import org.ivdnt.galahad.port.SourceLayerableFile
 import org.ivdnt.galahad.port.conllu.ConlluFile
 import org.ivdnt.galahad.port.tsv.export.TSVLayerMerger
 import java.io.File
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/LayerToTSVConverter.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/LayerToTSVConverter.kt
index cb87b06..36a278a 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/LayerToTSVConverter.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/LayerToTSVConverter.kt
@@ -18,8 +18,8 @@ class LayerToTSVConverter(
         outputStream.write("word\tlemma\tpos\n".encodeToByteArray()) // 'word' is the blacklab default
         // Body
         result.terms.forEach {
-            // Note that this might have weird result for multi-wordform tokens
-            outputStream.write("${it.literals}\t${it.lemma}\t${it.pos}\n".encodeToByteArray())
+            // Explicitly non-null.
+            outputStream.write("${it.literals}\t${it.lemmaOrEmpty}\t${it.posOrEmpty}\n".encodeToByteArray())
         }
     }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/TSVLayerMerger.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/TSVLayerMerger.kt
index a268c3b..24666d0 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/TSVLayerMerger.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/tsv/export/TSVLayerMerger.kt
@@ -1,63 +1,63 @@
-package org.ivdnt.galahad.port.tsv.export
-
-import org.ivdnt.galahad.data.layer.Layer
-import org.ivdnt.galahad.port.DocumentTransformMetadata
-import org.ivdnt.galahad.port.LayerMerger
-import org.ivdnt.galahad.port.LayerTransformer
-import org.ivdnt.galahad.port.tsv.TSVFile
-import java.io.File
-import kotlin.io.path.createTempDirectory
-
-/**
- * Do not call directly. Use [TSVFile.merge] instead.
- */
-internal open class TSVLayerMerger(
-    open val sourceFile: TSVFile,
-    transformMetadata: DocumentTransformMetadata,
-) : LayerMerger<TSVFile>, LayerTransformer(transformMetadata) {
-    val layer = transformMetadata.layer
-    val outFile: File = createTempDirectory("teimerge").toFile().resolve(transformMetadata.document.name)
-    protected open val hasHeader: Boolean = true
-    /**
-     * Merge uploaded raw file with tagger layer. Headers indices are already determined by TSVFile.
-     * Read in per line, split on tabs, swap out pos & lemma and commit to new file
-     */
-    override fun merge(): TSVFile {
-        sourceFile.parse() // parse the sourceFile if needed.
-        parseByLine()
-        return TSVFile(outFile)
-    }
-
-    protected fun parseByLine() {
-        var termIndex = if (hasHeader) -1 else 0 // Start at -1 to take the header into account.
-        sourceFile.file.inputStream().bufferedReader().forEachLine { line ->
-            if (termIndex == -1) {
-                // Copy header to output & continue
-                outFile.appendText(line + "\n")
-                termIndex++
-            } else {
-                val columns = line.split("\t").toMutableList()
-                if (columns.size >= 3) {
-                    // Swap out pos & lemma, keep the rest.
-                    replaceColumns(columns, layer, termIndex)
-                    outFile.appendText(columns.joinToString("\t") + "\n")
-                    termIndex++
-                } else {
-                    // Output whatever was on that line. Presumably whitespace.
-                    outFile.appendText(line + "\n")
-                }
-            }
-        }
-    }
-
-    /*
-     * Replace the PoS and lemma values in their previously indexed columns.
-     */
-    protected open fun replaceColumns(
-        columns: MutableList<String>, layer: Layer,
-        termIndex: Int,
-    ) {
-        columns[sourceFile.posIndex!!] = layer.terms[termIndex].pos.toString()
-        columns[sourceFile.lemmaIndex!!] = layer.terms[termIndex].lemma.toString()
-    }
+package org.ivdnt.galahad.port.tsv.export
+
+import org.ivdnt.galahad.data.layer.Layer
+import org.ivdnt.galahad.port.DocumentTransformMetadata
+import org.ivdnt.galahad.port.LayerMerger
+import org.ivdnt.galahad.port.LayerTransformer
+import org.ivdnt.galahad.port.tsv.TSVFile
+import java.io.File
+import kotlin.io.path.createTempDirectory
+
+/**
+ * Do not call directly. Use [TSVFile.merge] instead.
+ */
+internal open class TSVLayerMerger(
+    open val sourceFile: TSVFile,
+    transformMetadata: DocumentTransformMetadata,
+) : LayerMerger<TSVFile>, LayerTransformer(transformMetadata) {
+    val layer = transformMetadata.layer
+    val outFile: File = createTempDirectory("teimerge").toFile().resolve(transformMetadata.document.name)
+    protected open val hasHeader: Boolean = true
+    /**
+     * Merge uploaded raw file with tagger layer. Headers indices are already determined by TSVFile.
+     * Read in per line, split on tabs, swap out pos & lemma and commit to new file
+     */
+    override fun merge(): TSVFile {
+        sourceFile.parse() // parse the sourceFile if needed.
+        parseByLine()
+        return TSVFile(outFile)
+    }
+
+    protected fun parseByLine() {
+        var termIndex = if (hasHeader) -1 else 0 // Start at -1 to take the header into account.
+        sourceFile.file.inputStream().bufferedReader().forEachLine { line ->
+            if (termIndex == -1) {
+                // Copy header to output & continue
+                outFile.appendText(line + "\n")
+                termIndex++
+            } else {
+                val columns = line.split("\t").toMutableList()
+                if (columns.size >= 3) {
+                    // Swap out pos & lemma, keep the rest.
+                    replaceColumns(columns, layer, termIndex)
+                    outFile.appendText(columns.joinToString("\t") + "\n")
+                    termIndex++
+                } else {
+                    // Output whatever was on that line. Presumably whitespace.
+                    outFile.appendText(line + "\n")
+                }
+            }
+        }
+    }
+
+    /*
+     * Replace the PoS and lemma values in their previously indexed columns.
+     */
+    protected open fun replaceColumns(
+        columns: MutableList<String>, layer: Layer,
+        termIndex: Int,
+    ) {
+        columns[sourceFile.posIndex!!] = layer.terms[termIndex].posOrEmpty
+        columns[sourceFile.lemmaIndex!!] = layer.terms[termIndex].lemmaOrEmpty
+    }
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/BLFXMLParser.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/BLFXMLParser.kt
index de5826d..9f084a1 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/BLFXMLParser.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/BLFXMLParser.kt
@@ -1,10 +1,10 @@
 package org.ivdnt.galahad.port.xml
 
 import org.ivdnt.galahad.data.document.DocumentFormat
+import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.layer.Term
 import org.ivdnt.galahad.data.layer.WordForm
-import org.ivdnt.galahad.data.document.SOURCE_LAYER_NAME
 import org.ivdnt.galahad.port.BLFXML
 import org.ivdnt.galahad.util.getXmlBuilder
 import org.w3c.dom.Document
@@ -33,8 +33,8 @@ fun Node.tagName(): String? {
 }
 
 /**
- * Should the text text inside this node be interpreted as source text?
- * Asssumes we are already inside of a text container e.g. <text>
+ * Should the text inside this node be interpreted as source text?
+ * Assumes we are already inside a text container e.g. <text>
  */
 private fun Node.isTextable(): Boolean {
     if( this.tagName() == "note" && this.attributes.getNamedItem("type")?.textContent == "editorial" ) {
@@ -210,7 +210,7 @@ class BLFXMLParser (
 
 
 
-    private fun addPlaintext( literal: String ) {
+    private fun addPlaintext(literal: String) {
         plainTextOutputStream.write( literal.toByteArray() )
         offset += literal.length
 
@@ -277,20 +277,32 @@ class BLFXMLParser (
     }
 
     private fun handleWordOrPunctNode( node: Node ) {
-        // custom node handling
-        nodeHandler(node, offset, xmlDocument)
+        // Handle cases like <w>a</w><w>b</w> -> "a b" (add space in plaintext)
+        val needsSpacing = node.tagName() == "w" && plaintextTail.isNotBlank() && !Regex("""\s$""").containsMatchIn(plaintextTail)
+        val spaceOffset = if (needsSpacing) 1 else 0
+        val trueWordOffset = offset + spaceOffset
+
+        // Handle merging
+        nodeHandler(node, trueWordOffset, xmlDocument)
 
+        // Extraction
         val literal = literalExtractor(node).trim() // wordPathExpression.evaluate( node )
         val lem = lemmaExtractor(node) // lemPathExpression.evaluate( node )
         val pos = posExtractor(node) // posPathExpression.evaluate( node )
         val id = idExtractor(node)
 
-        val wordForm = WordForm( literal, offset, literal.length, id ?: "no-id" )
-        sourceLayer.wordForms.add( wordForm )
-        addPlaintext(literal)
-
+        // Add the word to the source layer
+        val wordForm = WordForm(literal, trueWordOffset, literal.length, id ?: "no-id" )
         val term = Term(lem, pos, mutableListOf(wordForm))
-        sourceLayer.terms.add( term )
+        sourceLayer.wordForms.add(wordForm)
+        sourceLayer.terms.add(term)
+
+        // Add the word to the plaintext
+        var text = literal.trim()
+        if (needsSpacing) {
+            text = " $text"
+        }
+        addPlaintext(text)
     }
 
     fun xmlToString(pretty: Boolean): String {
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLFile.kt
index 8c4711a..1615223 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLFile.kt
@@ -1,9 +1,7 @@
 package org.ivdnt.galahad.port.xml
 
 import org.ivdnt.galahad.port.InternalFile
-import org.xml.sax.Attributes
 import java.io.File
-import java.util.*
 
 abstract class XMLFile(
     final override val file: File,
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLMetadata.kt b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLMetadata.kt
index c73c98b..085cc85 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLMetadata.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/port/xml/XMLMetadata.kt
@@ -29,7 +29,7 @@ open class XMLMetadata(
     }
 
     /**
-     * Add a tag to [this] with [name], [textContent], and optional [attrValue] and [targetAttr].
+     * Add a tag to [this] with [name], [textContent], and optional [attrValue]
      * Defaults to writing attribute @type.
      */
     protected fun Node.createChild(
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/taggers/TaggersController.kt b/server/src/main/kotlin/org/ivdnt/galahad/taggers/TaggersController.kt
index a90dbb6..28dbf6a 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/taggers/TaggersController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/taggers/TaggersController.kt
@@ -4,25 +4,25 @@ import com.beust.klaxon.JsonObject
 import com.beust.klaxon.Parser
 import com.beust.klaxon.Parser.Companion.default
 import com.fasterxml.jackson.annotation.JsonProperty
+import jakarta.servlet.http.HttpServletRequest
+import jakarta.servlet.http.HttpServletResponse
 import org.apache.logging.log4j.kotlin.Logging
 import org.ivdnt.galahad.app.TAGGERS_URL
 import org.ivdnt.galahad.app.TAGGER_HEALTH_URL
 import org.ivdnt.galahad.app.TAGGER_URL
 import org.springframework.beans.factory.annotation.Autowired
+import org.springframework.http.HttpMethod
 import org.springframework.web.bind.annotation.CrossOrigin
 import org.springframework.web.bind.annotation.GetMapping
 import org.springframework.web.bind.annotation.PathVariable
 import org.springframework.web.bind.annotation.RestController
+import org.springframework.web.client.RestTemplate
+import org.springframework.web.util.UriComponentsBuilder
 import java.net.URI
+import java.net.URL
 import java.net.http.HttpClient
 import java.net.http.HttpRequest
 import java.net.http.HttpResponse
-import jakarta.servlet.http.HttpServletRequest
-import jakarta.servlet.http.HttpServletResponse
-import org.springframework.http.HttpMethod
-import org.springframework.web.client.RestTemplate
-import org.springframework.web.util.UriComponentsBuilder
-import java.net.URL
 
 @RestController
 class TaggersController : Logging {
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/tagset/TagsetController.kt b/server/src/main/kotlin/org/ivdnt/galahad/tagset/TagsetController.kt
index dc0d232..bade02a 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/tagset/TagsetController.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/tagset/TagsetController.kt
@@ -3,7 +3,10 @@ package org.ivdnt.galahad.tagset
 import org.apache.logging.log4j.kotlin.Logging
 import org.ivdnt.galahad.app.TAGSETS_URL
 import org.springframework.http.HttpStatus
-import org.springframework.web.bind.annotation.*
+import org.springframework.web.bind.annotation.CrossOrigin
+import org.springframework.web.bind.annotation.GetMapping
+import org.springframework.web.bind.annotation.PathVariable
+import org.springframework.web.bind.annotation.RestController
 import org.springframework.web.server.ResponseStatusException
 
 @RestController
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/util/ResourceUtil.kt b/server/src/main/kotlin/org/ivdnt/galahad/util/ResourceUtil.kt
index d94624e..4c8644c 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/util/ResourceUtil.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/util/ResourceUtil.kt
@@ -1,6 +1,7 @@
-package org.ivdnt.galahad.util
-
-import java.io.InputStream
-
-fun getResourceStream(path: String): InputStream? =
-    object {}.javaClass.classLoader.getResourceAsStream(path)
+package org.ivdnt.galahad.util
+
+import java.io.InputStream
+
+/** Get a resource from src/main/resources. */
+fun getResourceStream(path: String): InputStream? =
+    object {}.javaClass.classLoader.getResourceAsStream(path)
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/util/StringExtensions.kt b/server/src/main/kotlin/org/ivdnt/galahad/util/StringExtensions.kt
index 0bcd00a..6a52a48 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/util/StringExtensions.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/util/StringExtensions.kt
@@ -9,6 +9,7 @@ fun String.matchesUpTo(textToMatch: String): Int {
     return matchingIndex
 }
 
+/** Same as toString, unless the caller is null or the resulting string is empty: then we use [default]. */
 fun Any?.toNonEmptyString(default: String): String {
     return if (this == null || this.toString().isEmpty()) default else this.toString()
 }
\ No newline at end of file
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/util/UTF8Util.kt b/server/src/main/kotlin/org/ivdnt/galahad/util/UTF8Util.kt
index 8bb3d69..cbdd4d7 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/util/UTF8Util.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/util/UTF8Util.kt
@@ -1,7 +1,7 @@
 package org.ivdnt.galahad.util
 
-import java.net.URLEncoder
 import jakarta.servlet.http.HttpServletResponse
+import java.net.URLEncoder
 
 /**
  * A valid filename for windows and linux. Exceptions like COM1 still exist.
diff --git a/server/src/main/kotlin/org/ivdnt/galahad/util/ZipFile.kt b/server/src/main/kotlin/org/ivdnt/galahad/util/ZipFile.kt
index 9209b64..9e80cce 100644
--- a/server/src/main/kotlin/org/ivdnt/galahad/util/ZipFile.kt
+++ b/server/src/main/kotlin/org/ivdnt/galahad/util/ZipFile.kt
@@ -1,6 +1,9 @@
 package org.ivdnt.galahad.util
 
-import java.io.*
+import java.io.BufferedOutputStream
+import java.io.File
+import java.io.FileOutputStream
+import java.io.OutputStream
 import java.util.zip.ZipEntry
 import java.util.zip.ZipOutputStream
 
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/data/document/DocumentTest.kt b/server/src/test/kotlin/org/ivdnt/galahad/data/document/DocumentTest.kt
index 766e5c4..6bad97a 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/data/document/DocumentTest.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/data/document/DocumentTest.kt
@@ -91,7 +91,7 @@ class DocumentTest {
             // Convert to each other format
             for (formatTo in DocumentFormat.entries) {
                 val meta = DocumentTransformMetadata(
-                    corpus, job, doc, User("testUser")
+                    corpus, job, doc, User("testUser"), formatTo
                 )
                 when (formatTo) {
                     // Skip the unsupported
@@ -107,7 +107,7 @@ class DocumentTest {
                         val result: File = doc.generateAs(formatTo, meta)
                         val expected: File = Resource.get("all-formats/output/from-$formatFrom-to-$formatTo.${formatTo.extension}")
                         val test = TestResult(expected.readText(), result.readText())
-                        test.ignoreDate().ignoreUUID().result()
+                        test.ignoreDate().ignoreUUID().ignoreTrailingWhiteSpaces().result()
                     }
                 }
             }
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/port/Util.kt b/server/src/test/kotlin/org/ivdnt/galahad/port/Util.kt
index b219f24..670934d 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/port/Util.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/port/Util.kt
@@ -9,6 +9,7 @@ import org.ivdnt.galahad.app.User
 import org.ivdnt.galahad.data.corpus.Corpus
 import org.ivdnt.galahad.data.corpus.MutableCorpusMetadata
 import org.ivdnt.galahad.data.document.Document
+import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.data.layer.Term
 import org.ivdnt.galahad.data.layer.WordForm
@@ -51,7 +52,7 @@ fun getJsonMapper(): ObjectMapper {
 
 fun createCorpus(workdir: File? = null, isDataset: Boolean = false, isAdmin: Boolean = false): Corpus {
     val parent = workdir ?: createTempDirectory().toFile()
-    val corpus = Corpus(parent.resolve(UUID.randomUUID().toString()), User("you"))
+    val corpus = Corpus(parent.resolve(UUID.randomUUID().toString()))
     corpus.updateMetadata(
         MutableCorpusMetadata(
             "you",
@@ -189,16 +190,16 @@ class DocTestBuilder(
     /** The file extension is relevant, otherwise conversion will fail */
     fun getDummyTransformMetadata(
         layer: Layer,
-        ext: String? = null,
+        format: DocumentFormat,
         file: File? = null,
     ): DocumentTransformMetadata {
-        val file = file ?: createTempDirectory().toFile().resolve("dummy.$ext")
+        val file = file ?: createTempDirectory().toFile().resolve("dummy.${format.extension}")
         file.createNewFile()
         val docName = corpus.documents.create(file)
         val job = corpus.jobs.createOrThrow(TestConfig.TAGGER_NAME)
         job.document(docName).setResult(layer)
         return DocumentTransformMetadata(
-            corpus, job, corpus.documents.readOrThrow(docName), User("testUser")
+            corpus, job, corpus.documents.readOrThrow(docName), User("testUser"), format
         )
     }
 
@@ -206,7 +207,7 @@ class DocTestBuilder(
 
     fun convertToTSV(layer: Layer): TestResult {
         val exporter = LayerToTSVConverter(
-            getDummyTransformMetadata(layer, "tsv")
+            getDummyTransformMetadata(layer, DocumentFormat.Tsv)
         )
         val result = exporter.convertToFileNamed("test")
         return got(result.readText())
@@ -217,7 +218,7 @@ class DocTestBuilder(
     }
 
     fun mergeTSV(file: File, layer: Layer): TestResult {
-        val transformMetadata = getDummyTransformMetadata(layer, file = file)
+        val transformMetadata = getDummyTransformMetadata(layer, DocumentFormat.Tsv, file)
         val result: TSVFile = TSVFile(file).merge(transformMetadata)
         return got(result.file.readText())
     }
@@ -226,7 +227,7 @@ class DocTestBuilder(
 
     fun convertToConllu(layer: Layer): TestResult {
         val exporter = LayerToConlluConverter(
-            getDummyTransformMetadata(layer, "conllu")
+            getDummyTransformMetadata(layer, DocumentFormat.Conllu)
         )
         val result = exporter.convertToFileNamed("test")
         return got(result.readText())
@@ -237,7 +238,7 @@ class DocTestBuilder(
     }
 
     fun mergeConllu(file: File, layer: Layer): TestResult {
-        val transformMetadata = getDummyTransformMetadata(layer, file = file)
+        val transformMetadata = getDummyTransformMetadata(layer, DocumentFormat.Conllu, file)
         val result: ConlluFile = ConlluFile(file).merge(transformMetadata)
         return got(result.file.readText())
     }
@@ -246,7 +247,7 @@ class DocTestBuilder(
 
     fun convertToNaf(file: File, layer: Layer): TestResult {
         val exporter = LayerToNAFConverter(
-            getDummyTransformMetadata(layer, file = file)
+            getDummyTransformMetadata(layer, DocumentFormat.Naf, file)
         )
         val result = exporter.convertToFileNamed("test")
         return got(result.readText())
@@ -256,28 +257,26 @@ class DocTestBuilder(
 
     fun convertToFolia(file: File, layer: Layer): TestResult {
         val exporter = LayerToFoliaConverter(
-            getDummyTransformMetadata(layer, file = file)
+            getDummyTransformMetadata(layer, DocumentFormat.Folia, file)
         )
         val result = exporter.convertToFileNamed("test")
         return got(result.readText())
     }
 
     fun mergeFolia(file: File, layer: Layer): TestResult {
-        val transformMetadata = getDummyTransformMetadata(layer, file = file)
+        val transformMetadata = getDummyTransformMetadata(layer, DocumentFormat.Folia, file)
         val result: FoliaFile = FoliaFile(file).merge(transformMetadata)
         return got(result.file.readText())
     }
 
     // TEI
 
-    fun convertToTEI(teiFile: File, layer: Layer): TestResult {
-        val docName = corpus.documents.create(teiFile)
+    fun convertToTEI(file: File, layer: Layer): TestResult {
+        val docName = corpus.documents.create(file)
         val job = corpus.jobs.createOrThrow(TestConfig.TAGGER_NAME)
         job.document(docName).setResult(layer)
         val exporter = LayerToTEIConverter(
-            DocumentTransformMetadata(
-                corpus, job, corpus.documents.readOrThrow(docName), User("testUser")
-            )
+            getDummyTransformMetadata(layer, DocumentFormat.TeiP5, file)
         )
 
         val result = exporter.convertToFileNamed("tst")
@@ -289,7 +288,7 @@ class DocTestBuilder(
     }
 
     fun mergeTEI(file: File, layer: Layer): TestResult {
-        val transformMetadata = getDummyTransformMetadata(layer, file = file)
+        val transformMetadata = getDummyTransformMetadata(layer, DocumentFormat.TeiP5, file)
         val result: TEIFile = TEIFile(file).merge(transformMetadata)
         return got(result.file.readText())
     }
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/port/naf/NafExportTest.kt b/server/src/test/kotlin/org/ivdnt/galahad/port/naf/NafExportTest.kt
index e99a21e..9637f40 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/port/naf/NafExportTest.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/port/naf/NafExportTest.kt
@@ -1,6 +1,7 @@
 package org.ivdnt.galahad.port.naf
 
 import org.ivdnt.galahad.data.corpus.Corpus
+import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.port.DocTest
 import org.ivdnt.galahad.port.LayerBuilder
 import org.ivdnt.galahad.port.Resource
@@ -38,7 +39,7 @@ class NafExportTest {
     @Test
     fun `Merge throws`() {
         val layer = LayerBuilder().loadDummies(1).build()
-        val meta = DocTest.builder(corpus).getDummyTransformMetadata(layer, "txt")
+        val meta = DocTest.builder(corpus).getDummyTransformMetadata(layer, DocumentFormat.Txt)
         assertThrows(Exception::class.java) {
             Resource.getDoc("naf/import/input.naf.xml").merge(meta)
         }
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/port/plain/PlainFileTest.kt b/server/src/test/kotlin/org/ivdnt/galahad/port/plain/PlainFileTest.kt
index a49a159..0d471d2 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/port/plain/PlainFileTest.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/port/plain/PlainFileTest.kt
@@ -1,6 +1,7 @@
 package org.ivdnt.galahad.port.plain
 
 import org.ivdnt.galahad.data.corpus.Corpus
+import org.ivdnt.galahad.data.document.DocumentFormat
 import org.ivdnt.galahad.data.layer.Layer
 import org.ivdnt.galahad.port.DocTestBuilder
 import org.ivdnt.galahad.port.Resource
@@ -30,7 +31,7 @@ internal class PlainFileTest {
         val builder = DocTestBuilder(corpus)
         val file = PlainFile(Resource.get("txt/input.txt"))
         assertThrows(Exception::class.java) {
-            file.merge(builder.getDummyTransformMetadata(Layer.EMPTY, "txt"))
+            file.merge(builder.getDummyTransformMetadata(Layer.EMPTY, DocumentFormat.Txt))
         }
     }
 }
\ No newline at end of file
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIExportTest.kt b/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIExportTest.kt
index d729d06..0c6daae 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIExportTest.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIExportTest.kt
@@ -22,6 +22,9 @@ internal class TEIExportTest {
 
     @Test
     fun `Merge pie-tdn result with heavily twined tei`() {
+        val file = TEIFile(Resource.get("tei/twine/twine.input.xml"))
+        assertPlainText("tei/twine", file)
+
         val plaintext: String = Resource.get("tei/twine/plaintext.txt").readText()
         val layer = LayerBuilder()
             .loadLayerFromTSV("tei/twine/pie-tdn.tsv", plaintext)
@@ -37,6 +40,9 @@ internal class TEIExportTest {
 
     @Test
     fun `Merge a pie-tdn layer with a tei file that only contains plaintext`() {
+        val file = TEIFile(Resource.get("tei/brieven/input.tei.xml"))
+        assertPlainText("tei/brieven", file)
+
         val plaintext: String = Resource.get("tei/brieven/plaintext.txt").readText()
         val layer = LayerBuilder()
             .loadLayerFromTSV("tei/brieven/pie.tsv", plaintext)
@@ -152,7 +158,8 @@ internal class TEIExportTest {
                     corpus = corpus,
                     job = corpus.jobs.readOrThrow(jobName),
                     document = corpus.documents.readOrThrow(docName),
-                    user = User("test-user")
+                    user = User("test-user"),
+                    targetFormat = DocumentFormat.TeiP5
                 )
         ) }
 
@@ -170,7 +177,8 @@ internal class TEIExportTest {
                     corpus = corpus,
                     job = corpus.jobs.readOrThrow(jobName),
                     document = corpus.documents.readOrThrow(teiUploadedFileName),
-                    user = User("test-user")
+                    user = User("test-user"),
+                    targetFormat = DocumentFormat.TeiP5
                 )
             )
         }
diff --git a/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIImportTest.kt b/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIImportTest.kt
index bb441c7..7ef9429 100644
--- a/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIImportTest.kt
+++ b/server/src/test/kotlin/org/ivdnt/galahad/port/tei/TEIImportTest.kt
@@ -13,7 +13,7 @@ internal class TEIImportTest {
         @Test
         fun `Multiple text elements`() {
             val teiFile = TEIFile(Resource.get("tei/dummies/multipletextelements.xml"))
-            assertEquals("text1\ntext2text3", teiFile.plainTextReader().readText().trim())
+            assertEquals("text1\ntext2 text3", teiFile.plainTextReader().readText().trim())
         }
 
         @Test
@@ -40,6 +40,13 @@ internal class TEIImportTest {
             // Has no source layer
             assertPlainText("tei/brieven", file)
         }
+
+        @Test
+        fun `Import TEI with w-tags without spaces in between`() {
+            val file = TEIFile(Resource.get("tei/nospaces/input.tei.xml"))
+            assertEquals("a a a", file.plainTextReader().readText().trim())
+
+        }
     }
 
     @Nested
diff --git a/server/src/test/resources/all-formats/input/input.folia.xml b/server/src/test/resources/all-formats/input/input.folia.xml
index b5d232c..dc0cbf3 100644
--- a/server/src/test/resources/all-formats/input/input.folia.xml
+++ b/server/src/test/resources/all-formats/input/input.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w15">
                 <t>"</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w16">
@@ -116,12 +116,12 @@
             </w>
             <w xml:id="id_input.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w21">
                 <t>"</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
diff --git a/server/src/test/resources/all-formats/input/input.tsv b/server/src/test/resources/all-formats/input/input.tsv
index 10def0d..38fa3a7 100644
--- a/server/src/test/resources/all-formats/input/input.tsv
+++ b/server/src/test/resources/all-formats/input/input.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
\ No newline at end of file
+?		LET
+"		LET
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-Conllu-to-Folia.folia.xml b/server/src/test/resources/all-formats/output/from-Conllu-to-Folia.folia.xml
index 155bcef..95a563f 100644
--- a/server/src/test/resources/all-formats/output/from-Conllu-to-Folia.folia.xml
+++ b/server/src/test/resources/all-formats/output/from-Conllu-to-Folia.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w15">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w16">
@@ -116,14 +116,14 @@
             </w>
             <w xml:id="id_input.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w21">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
     </text>
-</FoLiA>
+</FoLiA>
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-Conllu-to-Tsv.tsv b/server/src/test/resources/all-formats/output/from-Conllu-to-Tsv.tsv
index d75894e..3764c09 100644
--- a/server/src/test/resources/all-formats/output/from-Conllu-to-Tsv.tsv
+++ b/server/src/test/resources/all-formats/output/from-Conllu-to-Tsv.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
+?		LET
+"		LET
diff --git a/server/src/test/resources/all-formats/output/from-Folia-to-Tsv.tsv b/server/src/test/resources/all-formats/output/from-Folia-to-Tsv.tsv
index d75894e..3764c09 100644
--- a/server/src/test/resources/all-formats/output/from-Folia-to-Tsv.tsv
+++ b/server/src/test/resources/all-formats/output/from-Folia-to-Tsv.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
+?		LET
+"		LET
diff --git a/server/src/test/resources/all-formats/output/from-Naf-to-Folia.folia.xml b/server/src/test/resources/all-formats/output/from-Naf-to-Folia.folia.xml
index 162375f..54f8f29 100644
--- a/server/src/test/resources/all-formats/output/from-Naf-to-Folia.folia.xml
+++ b/server/src/test/resources/all-formats/output/from-Naf-to-Folia.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input_naf.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input_naf.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input_naf.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w15">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w16">
@@ -116,14 +116,14 @@
             </w>
             <w xml:id="id_input_naf.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_naf.p1.w21">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
     </text>
-</FoLiA>
+</FoLiA>
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-Naf-to-Tsv.tsv b/server/src/test/resources/all-formats/output/from-Naf-to-Tsv.tsv
index d75894e..3764c09 100644
--- a/server/src/test/resources/all-formats/output/from-Naf-to-Tsv.tsv
+++ b/server/src/test/resources/all-formats/output/from-Naf-to-Tsv.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
+?		LET
+"		LET
diff --git a/server/src/test/resources/all-formats/output/from-TeiP5-to-Folia.folia.xml b/server/src/test/resources/all-formats/output/from-TeiP5-to-Folia.folia.xml
index 63ffb00..6f28cee 100644
--- a/server/src/test/resources/all-formats/output/from-TeiP5-to-Folia.folia.xml
+++ b/server/src/test/resources/all-formats/output/from-TeiP5-to-Folia.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input_tei.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input_tei.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input_tei.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w15">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w16">
@@ -116,14 +116,14 @@
             </w>
             <w xml:id="id_input_tei.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input_tei.p1.w21">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
     </text>
-</FoLiA>
+</FoLiA>
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-TeiP5-to-Tsv.tsv b/server/src/test/resources/all-formats/output/from-TeiP5-to-Tsv.tsv
index d75894e..3764c09 100644
--- a/server/src/test/resources/all-formats/output/from-TeiP5-to-Tsv.tsv
+++ b/server/src/test/resources/all-formats/output/from-TeiP5-to-Tsv.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
+?		LET
+"		LET
diff --git a/server/src/test/resources/all-formats/output/from-Tsv-to-Folia.folia.xml b/server/src/test/resources/all-formats/output/from-Tsv-to-Folia.folia.xml
index 155bcef..95a563f 100644
--- a/server/src/test/resources/all-formats/output/from-Tsv-to-Folia.folia.xml
+++ b/server/src/test/resources/all-formats/output/from-Tsv-to-Folia.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w15">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w16">
@@ -116,14 +116,14 @@
             </w>
             <w xml:id="id_input.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w21">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
     </text>
-</FoLiA>
+</FoLiA>
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-Txt-to-Folia.folia.xml b/server/src/test/resources/all-formats/output/from-Txt-to-Folia.folia.xml
index 155bcef..95a563f 100644
--- a/server/src/test/resources/all-formats/output/from-Txt-to-Folia.folia.xml
+++ b/server/src/test/resources/all-formats/output/from-Txt-to-Folia.folia.xml
@@ -36,7 +36,7 @@
             </w>
             <w xml:id="id_input.p1.w4">
                 <t>,</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w5">
@@ -51,17 +51,17 @@
             </w>
             <w xml:id="id_input.p1.w7">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w8">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w9">
                 <t>.</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w10">
@@ -86,12 +86,12 @@
             </w>
             <w xml:id="id_input.p1.w14">
                 <t>:</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w15">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w16">
@@ -116,14 +116,14 @@
             </w>
             <w xml:id="id_input.p1.w20">
                 <t>?</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
             <w xml:id="id_input.p1.w21">
                 <t>&quot;</t>
-                <lemma class="null" processor="pie-tdn-all" set="pie-tdn-all"/>
+                <lemma class="" processor="pie-tdn-all" set="pie-tdn-all"/>
                 <pos class="LET" head="LET" processor="pie-tdn-all" set="pie-tdn-all"/>
             </w>
         </p>
     </text>
-</FoLiA>
+</FoLiA>
\ No newline at end of file
diff --git a/server/src/test/resources/all-formats/output/from-Txt-to-Tsv.tsv b/server/src/test/resources/all-formats/output/from-Txt-to-Tsv.tsv
index d75894e..3764c09 100644
--- a/server/src/test/resources/all-formats/output/from-Txt-to-Tsv.tsv
+++ b/server/src/test/resources/all-formats/output/from-Txt-to-Tsv.tsv
@@ -2,21 +2,21 @@ word	lemma	pos
 Hebban	hebben	VRB(finiteness=fin,tense=past)
 olla	olle	RES(type=for)
 vogala	vogala	NOU-C(number=sg)
-,	null	LET
+,		LET
 nestas	nestatis	RES(type=for)
 hagunnan	haguna	NOU-C(number=sg)
-.	null	LET
-.	null	LET
-.	null	LET
+.		LET
+.		LET
+.		LET
 Hinase	hinas	NOU-P
 ic	ik	PD(type=pers,position=free)
 ende	en	CONJ(type=coor)
 thu	huben	ADV(type=reg)
-:	null	LET
-"	null	LET
+:		LET
+"		LET
 uuat	uiteen	PD(type=w-p,position=free)
 unbidan	unibent	CONJ(type=sub)
 wi	wij	PD(type=pers,position=free)
 nu	nu	ADV(type=reg)
-?	null	LET
-"	null	LET
+?		LET
+"		LET
diff --git a/server/src/test/resources/folia/twine/merged-output.folia.xml b/server/src/test/resources/folia/twine/merged-output.folia.xml
index a95f5ff..331bd13 100644
--- a/server/src/test/resources/folia/twine/merged-output.folia.xml
+++ b/server/src/test/resources/folia/twine/merged-output.folia.xml
@@ -23,7 +23,7 @@
       <div class="div" xml:id="CLVN-1-1_71c1b8b5-a9c8-4c73-b214-5a926126eab3.text.1.body.1.div.1">
         <p class="p" xml:id="CLVN-1-1_71c1b8b5-a9c8-4c73-b214-5a926126eab3.text.1.body.1.div.1.p.1">
           <!-- Note the "be,<br />that": this should be parsed as "be,that". No space! Why? Because some documents contain words like "be<br/>kijken" -> "bekijken"-->
-          <w><t>To</t><lemma class="te" processor="path" set="path"/><pos class="ADP(type=pre)" head="ADP" processor="path" set="path"/></w> <w><t>ob<t-style class="italic">ey</t-style></t><lemma class="obedië" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w> <w><t>or</t><lemma class="or" processor="path" set="path"/><pos class="INT" head="INT" processor="path" set="path"/></w> <w><t>n<t-style class="bold">o</t-style>t</t><lemma class="not" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w> <w><t>to</t><lemma class="te" processor="path" set="path"/><pos class="ADP(type=pre)" head="ADP" processor="path" set="path"/></w> <w><t>be</t><lemma class="bedie" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w><w><t>,</t><lemma class="null" processor="path" set="path"/><pos class="LET" head="LET" processor="path" set="path"/></w><br/><w><t>that</t><lemma class="het+hat" processor="path" set="path"/><pos class="PD(type=d-p,subtype=art,position=prenom)+NOU-C(number=sg)" head="PD+NOU-C" processor="path" set="path"/></w> <w><t>is</t><lemma class="zijn" processor="path" set="path"/><pos class="VRB(finiteness=fin,tense=pres)" head="VRB" processor="path" set="path"/></w> <w><t>the</t><lemma class="the" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w><w><t><t-style class="bold">one</t-style></t><lemma class="one" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w><w><t><t-style class="bold"><t-style class="red">question</t-style></t-style></t><lemma class="question" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w> <w><t><t-style class="bold">that</t-style></t><lemma class="het+hat" processor="path" set="path"/><pos class="PD(type=d-p,subtype=art,position=prenom)+NOU-C(number=sg)" head="PD+NOU-C" processor="path" set="path"/></w> <w><t><t-style class="bold">y<t-style class="italic">re</t-style>main</t-style>ed</t><lemma class="jermina" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w><w><t>.</t><lemma class="null" processor="path" set="path"/><pos class="LET" head="LET" processor="path" set="path"/></w>
+          <w><t>To</t><lemma class="te" processor="path" set="path"/><pos class="ADP(type=pre)" head="ADP" processor="path" set="path"/></w> <w><t>ob<t-style class="italic">ey</t-style></t><lemma class="obedië" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w> <w><t>or</t><lemma class="or" processor="path" set="path"/><pos class="INT" head="INT" processor="path" set="path"/></w> <w><t>n<t-style class="bold">o</t-style>t</t><lemma class="not" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w> <w><t>to</t><lemma class="te" processor="path" set="path"/><pos class="ADP(type=pre)" head="ADP" processor="path" set="path"/></w> <w><t>be</t><lemma class="bedie" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w><w><t>,</t><lemma class="" processor="path" set="path"/><pos class="LET" head="LET" processor="path" set="path"/></w><br/><w><t>that</t><lemma class="het+hat" processor="path" set="path"/><pos class="PD(type=d-p,subtype=art,position=prenom)+NOU-C(number=sg)" head="PD+NOU-C" processor="path" set="path"/></w> <w><t>is</t><lemma class="zijn" processor="path" set="path"/><pos class="VRB(finiteness=fin,tense=pres)" head="VRB" processor="path" set="path"/></w> <w><t>the</t><lemma class="the" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w><w><t><t-style class="bold">one</t-style></t><lemma class="one" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w><w><t><t-style class="bold"><t-style class="red">question</t-style></t-style></t><lemma class="question" processor="path" set="path"/><pos class="RES(type=for)" head="RES" processor="path" set="path"/></w> <w><t><t-style class="bold">that</t-style></t><lemma class="het+hat" processor="path" set="path"/><pos class="PD(type=d-p,subtype=art,position=prenom)+NOU-C(number=sg)" head="PD+NOU-C" processor="path" set="path"/></w> <w><t><t-style class="bold">y<t-style class="italic">re</t-style>main</t-style>ed</t><lemma class="jermina" processor="path" set="path"/><pos class="NOU-C(number=sg)" head="NOU-C" processor="path" set="path"/></w><w><t>.</t><lemma class="" processor="path" set="path"/><pos class="LET" head="LET" processor="path" set="path"/></w>
           <lang class="nld"/>
         </p>
         <p class="p" xml:id="CLVN-1-1_71c1b8b5-a9c8-4c73-b214-5a926126eab3.text.1.body.1.div.1.p.2">
diff --git a/server/src/test/resources/tei/nospaces/input.tei.xml b/server/src/test/resources/tei/nospaces/input.tei.xml
new file mode 100644
index 0000000..4ac729d
--- /dev/null
+++ b/server/src/test/resources/tei/nospaces/input.tei.xml
@@ -0,0 +1,7 @@
+<TEI>
+    <text>
+        <p>
+            <w>a</w><w>a</w><w>a</w>
+        </p>
+    </text>
+</TEI>
\ No newline at end of file