TagTagWord2Vec.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploring Metadata Keyword Relationships with word2vec\n",
    "\n",
    "This notebook demonstrates using a pretrained word2vec model to explore how keywords are related to other words and other keywords. \n",
    "\n",
    "Examples of how to use pretrained models and code where used from https://github.com/3Top/word2vec-api\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook runs on the docker jupyter/datascience-notebook container with gensim additionally installed as below "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "pip install gensim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from gensim.models.word2vec import Word2Vec as w\n",
    "import gensim.models\n",
    "from gensim import utils, matutils"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading a pretrained word2vec model\n",
    "\n",
    "This model has been sourced http://nlp.stanford.edu/data/glove.6B.zip to use this you'll need to download it as it's too big to package alongside this notebook.\n",
    "\n",
    "To load it sucessfully it has been modified to add a '400000 300' (presumably specifying the number of words and vector size) to a new line above the rest of the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "wget http://nlp.stanford.edu/data/glove.6B.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "unzip -o 'glove.6B.zip'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "sed -i '1s/^/400000 300\\n/' glove.6B.300d.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash \n",
    "head -2 ./glove.6B.300d.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gensim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = gensim.models.KeyedVectors.load_word2vec_format('./glove.6B.300d.txt', binary=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keywords describing datasets\n",
    "\n",
    "A keywords.json file was produced from backend data harvested as part of populating the knowledge network http://kn.csiro.au"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('./keywords.json') as keywords_json_file:\n",
    "    raw_keywords = json.load(keywords_json_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Lots of datasets but only a few of them have keywords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "keywords_dict = [keyword for keyword in raw_keywords if 'keywords' in keyword.keys() and keyword['keywords'] != None]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "67695\n",
      "1822\n"
     ]
    }
   ],
   "source": [
    "print(len(raw_keywords))\n",
    "print(len(keywords_dict))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>_id</th>\n",
       "      <th>keywords</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59159e9b3dd97b0a4c165378</td>\n",
       "      <td>[entomology, insects, spiders, fauna]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>59159e9b3dd97b0a4c165379</td>\n",
       "      <td>[plants]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>59159e9b3dd97b0a4c16537b</td>\n",
       "      <td>[microbial]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>59159e9b3dd97b0a4c16537e</td>\n",
       "      <td>[plants]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59159e9b3dd97b0a4c16537f</td>\n",
       "      <td>[herbarium, herbaria, angiosperms, dicots, monocots, pteridophytes, mosses, liverworts, lichens, fungi, algae, plants]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>59159e9b3dd97b0a4c165380</td>\n",
       "      <td>[microbial]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>59159e9b3dd97b0a4c165381</td>\n",
       "      <td>[microbial]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>59159e9b3dd97b0a4c165382</td>\n",
       "      <td>[microbes]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>59159e9b3dd97b0a4c165383</td>\n",
       "      <td>[arachnids, spiders, myriapods, centipedes, millipedes, onychophorans, \"velvet, worms\", tardigrades, \"water, bears\", fauna]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>59159e9b3dd97b0a4c165384</td>\n",
       "      <td>[entomology, insects, hexapoda]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        _id  \\\n",
       "0  59159e9b3dd97b0a4c165378   \n",
       "1  59159e9b3dd97b0a4c165379   \n",
       "2  59159e9b3dd97b0a4c16537b   \n",
       "3  59159e9b3dd97b0a4c16537e   \n",
       "4  59159e9b3dd97b0a4c16537f   \n",
       "5  59159e9b3dd97b0a4c165380   \n",
       "6  59159e9b3dd97b0a4c165381   \n",
       "7  59159e9b3dd97b0a4c165382   \n",
       "8  59159e9b3dd97b0a4c165383   \n",
       "9  59159e9b3dd97b0a4c165384   \n",
       "\n",
       "                                                                                                                      keywords  \n",
       "0                                                                                        [entomology, insects, spiders, fauna]  \n",
       "1                                                                                                                     [plants]  \n",
       "2                                                                                                                  [microbial]  \n",
       "3                                                                                                                     [plants]  \n",
       "4       [herbarium, herbaria, angiosperms, dicots, monocots, pteridophytes, mosses, liverworts, lichens, fungi, algae, plants]  \n",
       "5                                                                                                                  [microbial]  \n",
       "6                                                                                                                  [microbial]  \n",
       "7                                                                                                                   [microbes]  \n",
       "8  [arachnids, spiders, myriapods, centipedes, millipedes, onychophorans, \"velvet, worms\", tardigrades, \"water, bears\", fauna]  \n",
       "9                                                                                              [entomology, insects, hexapoda]  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "pd.set_option('max_colwidth', 1000)\n",
    "\n",
    "df = pd.DataFrame.from_dict(keywords_dict)\n",
    "df[0:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keyword cleaning and considerations\n",
    "\n",
    "Keywords are flattened to a unique list. However there are a lot of keywords that aren't single words. These aren't handled currently. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "flat_keywords = [keyword.lower() for keyword_item in keywords_dict for keyword in keyword_item['keywords']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "unique_keywords = list(set(flat_keywords))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1812"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(unique_keywords)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Finding similar words\n",
    "\n",
    "Here we find the top 10 words similar to a keyword. Note that this isn't finding the most similar other keywords rather it is finding words from the original word to vec corpus (in this case wikipedia + gigawords), only a small number of which are exisiting keywords. Also note that not all keywords are in the corpus if a keyword isn't present an error is thrown."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['cattle',\n",
       " 'cloud reflectance',\n",
       " 'mgoh group content',\n",
       " 'biofilm',\n",
       " 'nipah virus',\n",
       " 'flinders catchment',\n",
       " 'climex',\n",
       " 'soil attributes',\n",
       " 'air humidity, plant',\n",
       " 'atmosphere winds',\n",
       " 'economic',\n",
       " 'pharmacokinetic',\n",
       " 'eukaryotes',\n",
       " 'tropical pastures',\n",
       " 'image analysis, cdna microarray analysis',\n",
       " 'bromoform ',\n",
       " 'biodiversity, generalised vegetation patterns',\n",
       " 'wireless sensor network',\n",
       " 'vegetation',\n",
       " 'wind, direction, lansdown, sensor',\n",
       " 'wave period',\n",
       " 'asparagus asparagoides',\n",
       " 'gear trials: in2015_e01',\n",
       " 'lauraceae',\n",
       " 'data-constrained modelling, x-ray ct, carbonate, calcite, dolomite, oil &amp',\n",
       " 'humus',\n",
       " 'aspect',\n",
       " 'pipeline',\n",
       " 'flower']"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_keywords[1:30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('flowers', 0.850063681602478),\n",
       " ('petals', 0.7753748893737793),\n",
       " ('floral', 0.7735286355018616),\n",
       " ('blossoms', 0.7563701272010803),\n",
       " ('fruit', 0.7506299018859863),\n",
       " ('flowering', 0.7505371570587158),\n",
       " ('roses', 0.7504088282585144),\n",
       " ('buds', 0.745373010635376),\n",
       " ('blooms', 0.7371509671211243),\n",
       " ('lavender', 0.7359495162963867)]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar_cosmul(positive='flower',topn=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>cattle</th>\n",
       "      <td>(livestock, 0.8918353915214539)</td>\n",
       "      <td>(sheep, 0.8761770725250244)</td>\n",
       "      <td>(cows, 0.8659452199935913)</td>\n",
       "      <td>(herds, 0.8254205584526062)</td>\n",
       "      <td>(cow, 0.8124057650566101)</td>\n",
       "      <td>(herd, 0.8111633658409119)</td>\n",
       "      <td>(goats, 0.8096607327461243)</td>\n",
       "      <td>(pigs, 0.7950717210769653)</td>\n",
       "      <td>(hogs, 0.7912872433662415)</td>\n",
       "      <td>(ranchers, 0.7834054231643677)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>cloud reflectance</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mgoh group content</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>biofilm</th>\n",
       "      <td>(biofilms, 0.7793806791305542)</td>\n",
       "      <td>(microbial, 0.7168845534324646)</td>\n",
       "      <td>(tumorigenesis, 0.711588442325592)</td>\n",
       "      <td>(overgrowth, 0.7092530727386475)</td>\n",
       "      <td>(microorganisms, 0.7048378586769104)</td>\n",
       "      <td>(phototrophic, 0.7012143731117249)</td>\n",
       "      <td>(filamentous, 0.7011398077011108)</td>\n",
       "      <td>(bacterial, 0.700860857963562)</td>\n",
       "      <td>(gingivitis, 0.696479082107544)</td>\n",
       "      <td>(epithelial, 0.6950613260269165)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>nipah virus</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>flinders catchment</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>climex</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>soil attributes</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>air humidity, plant</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>atmosphere winds</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>economic</th>\n",
       "      <td>(economy, 0.8744267225265503)</td>\n",
       "      <td>(growth, 0.8342223763465881)</td>\n",
       "      <td>(financial, 0.8243797421455383)</td>\n",
       "      <td>(economies, 0.8140730857849121)</td>\n",
       "      <td>(development, 0.8000568151473999)</td>\n",
       "      <td>(crisis, 0.7978229522705078)</td>\n",
       "      <td>(policy, 0.7938908934593201)</td>\n",
       "      <td>(monetary, 0.7867799401283264)</td>\n",
       "      <td>(recovery, 0.7843355536460876)</td>\n",
       "      <td>(cooperation, 0.7817569971084595)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>pharmacokinetic</th>\n",
       "      <td>(pharmacodynamic, 0.8009185194969177)</td>\n",
       "      <td>(pharmacokinetics, 0.7438419461250305)</td>\n",
       "      <td>(tolerability, 0.7250052690505981)</td>\n",
       "      <td>(physicochemical, 0.7232503294944763)</td>\n",
       "      <td>(rheological, 0.7101091146469116)</td>\n",
       "      <td>(morphologic, 0.7099518179893494)</td>\n",
       "      <td>(preclinical, 0.7092452645301819)</td>\n",
       "      <td>(pharmacological, 0.7050716280937195)</td>\n",
       "      <td>(forcings, 0.7033190131187439)</td>\n",
       "      <td>(cholesterol-lowering, 0.7028400301933289)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>eukaryotes</th>\n",
       "      <td>(prokaryotes, 0.8928068280220032)</td>\n",
       "      <td>(eukaryotic, 0.8583177328109741)</td>\n",
       "      <td>(archaea, 0.8358689546585083)</td>\n",
       "      <td>(multicellular, 0.8013555407524109)</td>\n",
       "      <td>(organelles, 0.7904345393180847)</td>\n",
       "      <td>(protists, 0.7789451479911804)</td>\n",
       "      <td>(mitochondria, 0.7780953645706177)</td>\n",
       "      <td>(vertebrates, 0.7777019739151001)</td>\n",
       "      <td>(organisms, 0.7762153148651123)</td>\n",
       "      <td>(mammalian, 0.7726430892944336)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tropical pastures</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>image analysis, cdna</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bromoform</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>biodiversity, genera</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>wireless sensor netw</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>vegetation</th>\n",
       "      <td>(grasses, 0.8359103798866272)</td>\n",
       "      <td>(habitats, 0.8210142254829407)</td>\n",
       "      <td>(forests, 0.816139280796051)</td>\n",
       "      <td>(shrubs, 0.8148589730262756)</td>\n",
       "      <td>(grassland, 0.8108927607536316)</td>\n",
       "      <td>(grasslands, 0.8104864954948425)</td>\n",
       "      <td>(habitat, 0.8080835938453674)</td>\n",
       "      <td>(soils, 0.7897382974624634)</td>\n",
       "      <td>(dense, 0.7884434461593628)</td>\n",
       "      <td>(ecosystems, 0.7876676917076111)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>wind, direction, lan</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>wave period</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>asparagus asparagoid</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gear trials: in2015_</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>lauraceae</th>\n",
       "      <td>(myrtaceae, 0.8181697726249695)</td>\n",
       "      <td>(annonaceae, 0.8012610077857971)</td>\n",
       "      <td>(anacardiaceae, 0.7908108830451965)</td>\n",
       "      <td>(apocynaceae, 0.7893630862236023)</td>\n",
       "      <td>(euphorbiaceae, 0.7842628955841064)</td>\n",
       "      <td>(rosaceae, 0.7839576601982117)</td>\n",
       "      <td>(moraceae, 0.7837344408035278)</td>\n",
       "      <td>(phyllanthaceae, 0.7814502120018005)</td>\n",
       "      <td>(meliaceae, 0.7806121110916138)</td>\n",
       "      <td>(theaceae, 0.7782191634178162)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>data-constrained mod</th>\n",
       "      <td>Word Not Found</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>humus</th>\n",
       "      <td>(compost, 0.7665944695472717)</td>\n",
       "      <td>(loamy, 0.7555372714996338)</td>\n",
       "      <td>(topsoil, 0.7548052668571472)</td>\n",
       "      <td>(loam, 0.7437748908996582)</td>\n",
       "      <td>(compacted, 0.7431917190551758)</td>\n",
       "      <td>(soils, 0.7316522598266602)</td>\n",
       "      <td>(composted, 0.7310082316398621)</td>\n",
       "      <td>(mulch, 0.729546070098877)</td>\n",
       "      <td>(subsoil, 0.7257762551307678)</td>\n",
       "      <td>(humic, 0.7250461578369141)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>aspect</th>\n",
       "      <td>(aspects, 0.868253767490387)</td>\n",
       "      <td>(perspective, 0.7785563468933105)</td>\n",
       "      <td>(facet, 0.7779170870780945)</td>\n",
       "      <td>(element, 0.7756810188293457)</td>\n",
       "      <td>(context, 0.7729584574699402)</td>\n",
       "      <td>(unique, 0.7608357071876526)</td>\n",
       "      <td>(interesting, 0.7564591765403748)</td>\n",
       "      <td>(particular, 0.7552378177642822)</td>\n",
       "      <td>(defining, 0.7537569403648376)</td>\n",
       "      <td>(important, 0.7527256011962891)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>pipeline</th>\n",
       "      <td>(pipelines, 0.8790407180786133)</td>\n",
       "      <td>(gas, 0.7858619689941406)</td>\n",
       "      <td>(oil, 0.7832066416740417)</td>\n",
       "      <td>(caspian, 0.765547513961792)</td>\n",
       "      <td>(ceyhan, 0.7645342350006104)</td>\n",
       "      <td>(refinery, 0.7623959183692932)</td>\n",
       "      <td>(btc, 0.7479247450828552)</td>\n",
       "      <td>(refineries, 0.7366435527801514)</td>\n",
       "      <td>(oilfields, 0.7358177304267883)</td>\n",
       "      <td>(turkmenistan, 0.7313023805618286)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>flower</th>\n",
       "      <td>(flowers, 0.850063681602478)</td>\n",
       "      <td>(petals, 0.7753748893737793)</td>\n",
       "      <td>(floral, 0.7735286355018616)</td>\n",
       "      <td>(blossoms, 0.7563701272010803)</td>\n",
       "      <td>(fruit, 0.7506299018859863)</td>\n",
       "      <td>(flowering, 0.7505371570587158)</td>\n",
       "      <td>(roses, 0.7504088282585144)</td>\n",
       "      <td>(buds, 0.745373010635376)</td>\n",
       "      <td>(blooms, 0.7371509671211243)</td>\n",
       "      <td>(lavender, 0.7359495162963867)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                          0  \\\n",
       "cattle                      (livestock, 0.8918353915214539)   \n",
       "cloud reflectance                            Word Not Found   \n",
       "mgoh group content                           Word Not Found   \n",
       "biofilm                      (biofilms, 0.7793806791305542)   \n",
       "nipah virus                                  Word Not Found   \n",
       "flinders catchment                           Word Not Found   \n",
       "climex                                       Word Not Found   \n",
       "soil attributes                              Word Not Found   \n",
       "air humidity, plant                          Word Not Found   \n",
       "atmosphere winds                             Word Not Found   \n",
       "economic                      (economy, 0.8744267225265503)   \n",
       "pharmacokinetic       (pharmacodynamic, 0.8009185194969177)   \n",
       "eukaryotes                (prokaryotes, 0.8928068280220032)   \n",
       "tropical pastures                            Word Not Found   \n",
       "image analysis, cdna                         Word Not Found   \n",
       "bromoform                                    Word Not Found   \n",
       "biodiversity, genera                         Word Not Found   \n",
       "wireless sensor netw                         Word Not Found   \n",
       "vegetation                    (grasses, 0.8359103798866272)   \n",
       "wind, direction, lan                         Word Not Found   \n",
       "wave period                                  Word Not Found   \n",
       "asparagus asparagoid                         Word Not Found   \n",
       "gear trials: in2015_                         Word Not Found   \n",
       "lauraceae                   (myrtaceae, 0.8181697726249695)   \n",
       "data-constrained mod                         Word Not Found   \n",
       "humus                         (compost, 0.7665944695472717)   \n",
       "aspect                         (aspects, 0.868253767490387)   \n",
       "pipeline                    (pipelines, 0.8790407180786133)   \n",
       "flower                         (flowers, 0.850063681602478)   \n",
       "\n",
       "                                                           1  \\\n",
       "cattle                           (sheep, 0.8761770725250244)   \n",
       "cloud reflectance                                       None   \n",
       "mgoh group content                                      None   \n",
       "biofilm                      (microbial, 0.7168845534324646)   \n",
       "nipah virus                                             None   \n",
       "flinders catchment                                      None   \n",
       "climex                                                  None   \n",
       "soil attributes                                         None   \n",
       "air humidity, plant                                     None   \n",
       "atmosphere winds                                        None   \n",
       "economic                        (growth, 0.8342223763465881)   \n",
       "pharmacokinetic       (pharmacokinetics, 0.7438419461250305)   \n",
       "eukaryotes                  (eukaryotic, 0.8583177328109741)   \n",
       "tropical pastures                                       None   \n",
       "image analysis, cdna                                    None   \n",
       "bromoform                                               None   \n",
       "biodiversity, genera                                    None   \n",
       "wireless sensor netw                                    None   \n",
       "vegetation                    (habitats, 0.8210142254829407)   \n",
       "wind, direction, lan                                    None   \n",
       "wave period                                             None   \n",
       "asparagus asparagoid                                    None   \n",
       "gear trials: in2015_                                    None   \n",
       "lauraceae                   (annonaceae, 0.8012610077857971)   \n",
       "data-constrained mod                                    None   \n",
       "humus                            (loamy, 0.7555372714996338)   \n",
       "aspect                     (perspective, 0.7785563468933105)   \n",
       "pipeline                           (gas, 0.7858619689941406)   \n",
       "flower                          (petals, 0.7753748893737793)   \n",
       "\n",
       "                                                        2  \\\n",
       "cattle                         (cows, 0.8659452199935913)   \n",
       "cloud reflectance                                    None   \n",
       "mgoh group content                                   None   \n",
       "biofilm                (tumorigenesis, 0.711588442325592)   \n",
       "nipah virus                                          None   \n",
       "flinders catchment                                   None   \n",
       "climex                                               None   \n",
       "soil attributes                                      None   \n",
       "air humidity, plant                                  None   \n",
       "atmosphere winds                                     None   \n",
       "economic                  (financial, 0.8243797421455383)   \n",
       "pharmacokinetic        (tolerability, 0.7250052690505981)   \n",
       "eukaryotes                  (archaea, 0.8358689546585083)   \n",
       "tropical pastures                                    None   \n",
       "image analysis, cdna                                 None   \n",
       "bromoform                                            None   \n",
       "biodiversity, genera                                 None   \n",
       "wireless sensor netw                                 None   \n",
       "vegetation                   (forests, 0.816139280796051)   \n",
       "wind, direction, lan                                 None   \n",
       "wave period                                          None   \n",
       "asparagus asparagoid                                 None   \n",
       "gear trials: in2015_                                 None   \n",
       "lauraceae             (anacardiaceae, 0.7908108830451965)   \n",
       "data-constrained mod                                 None   \n",
       "humus                       (topsoil, 0.7548052668571472)   \n",
       "aspect                        (facet, 0.7779170870780945)   \n",
       "pipeline                        (oil, 0.7832066416740417)   \n",
       "flower                       (floral, 0.7735286355018616)   \n",
       "\n",
       "                                                          3  \\\n",
       "cattle                          (herds, 0.8254205584526062)   \n",
       "cloud reflectance                                      None   \n",
       "mgoh group content                                     None   \n",
       "biofilm                    (overgrowth, 0.7092530727386475)   \n",
       "nipah virus                                            None   \n",
       "flinders catchment                                     None   \n",
       "climex                                                 None   \n",
       "soil attributes                                        None   \n",
       "air humidity, plant                                    None   \n",
       "atmosphere winds                                       None   \n",
       "economic                    (economies, 0.8140730857849121)   \n",
       "pharmacokinetic       (physicochemical, 0.7232503294944763)   \n",
       "eukaryotes              (multicellular, 0.8013555407524109)   \n",
       "tropical pastures                                      None   \n",
       "image analysis, cdna                                   None   \n",
       "bromoform                                              None   \n",
       "biodiversity, genera                                   None   \n",
       "wireless sensor netw                                   None   \n",
       "vegetation                     (shrubs, 0.8148589730262756)   \n",
       "wind, direction, lan                                   None   \n",
       "wave period                                            None   \n",
       "asparagus asparagoid                                   None   \n",
       "gear trials: in2015_                                   None   \n",
       "lauraceae                 (apocynaceae, 0.7893630862236023)   \n",
       "data-constrained mod                                   None   \n",
       "humus                            (loam, 0.7437748908996582)   \n",
       "aspect                        (element, 0.7756810188293457)   \n",
       "pipeline                       (caspian, 0.765547513961792)   \n",
       "flower                       (blossoms, 0.7563701272010803)   \n",
       "\n",
       "                                                         4  \\\n",
       "cattle                           (cow, 0.8124057650566101)   \n",
       "cloud reflectance                                     None   \n",
       "mgoh group content                                    None   \n",
       "biofilm               (microorganisms, 0.7048378586769104)   \n",
       "nipah virus                                           None   \n",
       "flinders catchment                                    None   \n",
       "climex                                                None   \n",
       "soil attributes                                       None   \n",
       "air humidity, plant                                   None   \n",
       "atmosphere winds                                      None   \n",
       "economic                 (development, 0.8000568151473999)   \n",
       "pharmacokinetic          (rheological, 0.7101091146469116)   \n",
       "eukaryotes                (organelles, 0.7904345393180847)   \n",
       "tropical pastures                                     None   \n",
       "image analysis, cdna                                  None   \n",
       "bromoform                                             None   \n",
       "biodiversity, genera                                  None   \n",
       "wireless sensor netw                                  None   \n",
       "vegetation                 (grassland, 0.8108927607536316)   \n",
       "wind, direction, lan                                  None   \n",
       "wave period                                           None   \n",
       "asparagus asparagoid                                  None   \n",
       "gear trials: in2015_                                  None   \n",
       "lauraceae              (euphorbiaceae, 0.7842628955841064)   \n",
       "data-constrained mod                                  None   \n",
       "humus                      (compacted, 0.7431917190551758)   \n",
       "aspect                       (context, 0.7729584574699402)   \n",
       "pipeline                      (ceyhan, 0.7645342350006104)   \n",
       "flower                         (fruit, 0.7506299018859863)   \n",
       "\n",
       "                                                       5  \\\n",
       "cattle                        (herd, 0.8111633658409119)   \n",
       "cloud reflectance                                   None   \n",
       "mgoh group content                                  None   \n",
       "biofilm               (phototrophic, 0.7012143731117249)   \n",
       "nipah virus                                         None   \n",
       "flinders catchment                                  None   \n",
       "climex                                              None   \n",
       "soil attributes                                     None   \n",
       "air humidity, plant                                 None   \n",
       "atmosphere winds                                    None   \n",
       "economic                    (crisis, 0.7978229522705078)   \n",
       "pharmacokinetic        (morphologic, 0.7099518179893494)   \n",
       "eukaryotes                (protists, 0.7789451479911804)   \n",
       "tropical pastures                                   None   \n",
       "image analysis, cdna                                None   \n",
       "bromoform                                           None   \n",
       "biodiversity, genera                                None   \n",
       "wireless sensor netw                                None   \n",
       "vegetation              (grasslands, 0.8104864954948425)   \n",
       "wind, direction, lan                                None   \n",
       "wave period                                         None   \n",
       "asparagus asparagoid                                None   \n",
       "gear trials: in2015_                                None   \n",
       "lauraceae                 (rosaceae, 0.7839576601982117)   \n",
       "data-constrained mod                                None   \n",
       "humus                        (soils, 0.7316522598266602)   \n",
       "aspect                      (unique, 0.7608357071876526)   \n",
       "pipeline                  (refinery, 0.7623959183692932)   \n",
       "flower                   (flowering, 0.7505371570587158)   \n",
       "\n",
       "                                                       6  \\\n",
       "cattle                       (goats, 0.8096607327461243)   \n",
       "cloud reflectance                                   None   \n",
       "mgoh group content                                  None   \n",
       "biofilm                (filamentous, 0.7011398077011108)   \n",
       "nipah virus                                         None   \n",
       "flinders catchment                                  None   \n",
       "climex                                              None   \n",
       "soil attributes                                     None   \n",
       "air humidity, plant                                 None   \n",
       "atmosphere winds                                    None   \n",
       "economic                    (policy, 0.7938908934593201)   \n",
       "pharmacokinetic        (preclinical, 0.7092452645301819)   \n",
       "eukaryotes            (mitochondria, 0.7780953645706177)   \n",
       "tropical pastures                                   None   \n",
       "image analysis, cdna                                None   \n",
       "bromoform                                           None   \n",
       "biodiversity, genera                                None   \n",
       "wireless sensor netw                                None   \n",
       "vegetation                 (habitat, 0.8080835938453674)   \n",
       "wind, direction, lan                                None   \n",
       "wave period                                         None   \n",
       "asparagus asparagoid                                None   \n",
       "gear trials: in2015_                                None   \n",
       "lauraceae                 (moraceae, 0.7837344408035278)   \n",
       "data-constrained mod                                None   \n",
       "humus                    (composted, 0.7310082316398621)   \n",
       "aspect                 (interesting, 0.7564591765403748)   \n",
       "pipeline                       (btc, 0.7479247450828552)   \n",
       "flower                       (roses, 0.7504088282585144)   \n",
       "\n",
       "                                                          7  \\\n",
       "cattle                           (pigs, 0.7950717210769653)   \n",
       "cloud reflectance                                      None   \n",
       "mgoh group content                                     None   \n",
       "biofilm                      (bacterial, 0.700860857963562)   \n",
       "nipah virus                                            None   \n",
       "flinders catchment                                     None   \n",
       "climex                                                 None   \n",
       "soil attributes                                        None   \n",
       "air humidity, plant                                    None   \n",
       "atmosphere winds                                       None   \n",
       "economic                     (monetary, 0.7867799401283264)   \n",
       "pharmacokinetic       (pharmacological, 0.7050716280937195)   \n",
       "eukaryotes                (vertebrates, 0.7777019739151001)   \n",
       "tropical pastures                                      None   \n",
       "image analysis, cdna                                   None   \n",
       "bromoform                                              None   \n",
       "biodiversity, genera                                   None   \n",
       "wireless sensor netw                                   None   \n",
       "vegetation                      (soils, 0.7897382974624634)   \n",
       "wind, direction, lan                                   None   \n",
       "wave period                                            None   \n",
       "asparagus asparagoid                                   None   \n",
       "gear trials: in2015_                                   None   \n",
       "lauraceae              (phyllanthaceae, 0.7814502120018005)   \n",
       "data-constrained mod                                   None   \n",
       "humus                            (mulch, 0.729546070098877)   \n",
       "aspect                     (particular, 0.7552378177642822)   \n",
       "pipeline                   (refineries, 0.7366435527801514)   \n",
       "flower                            (buds, 0.745373010635376)   \n",
       "\n",
       "                                                    8  \\\n",
       "cattle                     (hogs, 0.7912872433662415)   \n",
       "cloud reflectance                                None   \n",
       "mgoh group content                               None   \n",
       "biofilm               (gingivitis, 0.696479082107544)   \n",
       "nipah virus                                      None   \n",
       "flinders catchment                               None   \n",
       "climex                                           None   \n",
       "soil attributes                                  None   \n",
       "air humidity, plant                              None   \n",
       "atmosphere winds                                 None   \n",
       "economic               (recovery, 0.7843355536460876)   \n",
       "pharmacokinetic        (forcings, 0.7033190131187439)   \n",
       "eukaryotes            (organisms, 0.7762153148651123)   \n",
       "tropical pastures                                None   \n",
       "image analysis, cdna                             None   \n",
       "bromoform                                        None   \n",
       "biodiversity, genera                             None   \n",
       "wireless sensor netw                             None   \n",
       "vegetation                (dense, 0.7884434461593628)   \n",
       "wind, direction, lan                             None   \n",
       "wave period                                      None   \n",
       "asparagus asparagoid                             None   \n",
       "gear trials: in2015_                             None   \n",
       "lauraceae             (meliaceae, 0.7806121110916138)   \n",
       "data-constrained mod                             None   \n",
       "humus                   (subsoil, 0.7257762551307678)   \n",
       "aspect                 (defining, 0.7537569403648376)   \n",
       "pipeline              (oilfields, 0.7358177304267883)   \n",
       "flower                   (blooms, 0.7371509671211243)   \n",
       "\n",
       "                                                               9  \n",
       "cattle                            (ranchers, 0.7834054231643677)  \n",
       "cloud reflectance                                           None  \n",
       "mgoh group content                                          None  \n",
       "biofilm                         (epithelial, 0.6950613260269165)  \n",
       "nipah virus                                                 None  \n",
       "flinders catchment                                          None  \n",
       "climex                                                      None  \n",
       "soil attributes                                             None  \n",
       "air humidity, plant                                         None  \n",
       "atmosphere winds                                            None  \n",
       "economic                       (cooperation, 0.7817569971084595)  \n",
       "pharmacokinetic       (cholesterol-lowering, 0.7028400301933289)  \n",
       "eukaryotes                       (mammalian, 0.7726430892944336)  \n",
       "tropical pastures                                           None  \n",
       "image analysis, cdna                                        None  \n",
       "bromoform                                                   None  \n",
       "biodiversity, genera                                        None  \n",
       "wireless sensor netw                                        None  \n",
       "vegetation                      (ecosystems, 0.7876676917076111)  \n",
       "wind, direction, lan                                        None  \n",
       "wave period                                                 None  \n",
       "asparagus asparagoid                                        None  \n",
       "gear trials: in2015_                                        None  \n",
       "lauraceae                         (theaceae, 0.7782191634178162)  \n",
       "data-constrained mod                                        None  \n",
       "humus                                (humic, 0.7250461578369141)  \n",
       "aspect                           (important, 0.7527256011962891)  \n",
       "pipeline                      (turkmenistan, 0.7313023805618286)  \n",
       "flower                            (lavender, 0.7359495162963867)  "
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keyword_similarities = {}\n",
    "for keyword in unique_keywords[1:30]:\n",
    "    try:\n",
    "        keyword_similarities[keyword[0:20]] = model.most_similar_cosmul(positive=keyword,topn=10)\n",
    "    except:\n",
    "        keyword_similarities[keyword[0:20]] = [\"Word Not Found\"]\n",
    "        continue\n",
    "#print(keyword_similarities)\n",
    "pd.DataFrame.from_dict(keyword_similarities, orient=\"index\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Finding only similar existing keywords\n",
    "\n",
    "In some use cases we might want to only want to find existing keywords similar to other keywords. This might help a user select a new keyword based on other similar keywords to an already selected keyword. The best way to do this might be to reset the word2vec model to use the keywords as a corpus and then \"copy\" the weights from an existing model (because we can't easily train specifically on the small keyword corpus). An alternative is to \"brute force\" this by build ordered dictionaries so that each keyword is associated with an list of all other keywords ordered by similarity. Not implemented yet but might begin like the following"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Source Keyword: ocean\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Keyword</th>\n",
       "      <th>Similarity Score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>375</th>\n",
       "      <td>ocean</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>535</th>\n",
       "      <td>sea</td>\n",
       "      <td>0.684473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>297</th>\n",
       "      <td>oceans</td>\n",
       "      <td>0.643771</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>451</th>\n",
       "      <td>coast</td>\n",
       "      <td>0.577876</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>360</th>\n",
       "      <td>coastal</td>\n",
       "      <td>0.544094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>186</th>\n",
       "      <td>waves</td>\n",
       "      <td>0.497618</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>214</th>\n",
       "      <td>tidal</td>\n",
       "      <td>0.478499</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>72</th>\n",
       "      <td>water</td>\n",
       "      <td>0.476902</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>reefs</td>\n",
       "      <td>0.457699</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>595</th>\n",
       "      <td>tropical</td>\n",
       "      <td>0.44709</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>seabed</td>\n",
       "      <td>0.428529</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>253</th>\n",
       "      <td>fish</td>\n",
       "      <td>0.397474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>415</th>\n",
       "      <td>marine</td>\n",
       "      <td>0.395392</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>559</th>\n",
       "      <td>sand</td>\n",
       "      <td>0.395074</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>416</th>\n",
       "      <td>weather</td>\n",
       "      <td>0.393229</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>484</th>\n",
       "      <td>wind</td>\n",
       "      <td>0.382924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>533</th>\n",
       "      <td>continental</td>\n",
       "      <td>0.356296</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63</th>\n",
       "      <td>ecosystem</td>\n",
       "      <td>0.356034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>232</th>\n",
       "      <td>natural</td>\n",
       "      <td>0.353634</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>474</th>\n",
       "      <td>temperature</td>\n",
       "      <td>0.351932</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>395</th>\n",
       "      <td>atoll</td>\n",
       "      <td>0.348955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>309</th>\n",
       "      <td>salinity</td>\n",
       "      <td>0.346013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>356</th>\n",
       "      <td>sediment</td>\n",
       "      <td>0.342138</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>wildlife</td>\n",
       "      <td>0.341011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>488</th>\n",
       "      <td>estuary</td>\n",
       "      <td>0.338248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>156</th>\n",
       "      <td>wave</td>\n",
       "      <td>0.331474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>location</td>\n",
       "      <td>0.33124</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>208</th>\n",
       "      <td>sediments</td>\n",
       "      <td>0.32903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>306</th>\n",
       "      <td>environment</td>\n",
       "      <td>0.328656</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>404</th>\n",
       "      <td>boundary</td>\n",
       "      <td>0.327967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>barley</td>\n",
       "      <td>-0.0190345</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>466</th>\n",
       "      <td>maize</td>\n",
       "      <td>-0.018247</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>295</th>\n",
       "      <td>cereal</td>\n",
       "      <td>-0.0182227</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>579</th>\n",
       "      <td>ferns</td>\n",
       "      <td>-0.0181958</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>517</th>\n",
       "      <td>workflow</td>\n",
       "      <td>-0.0171294</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>479</th>\n",
       "      <td>low-cost</td>\n",
       "      <td>-0.015994</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>576</th>\n",
       "      <td>fluorescence</td>\n",
       "      <td>-0.0154273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>221</th>\n",
       "      <td>myriapods</td>\n",
       "      <td>-0.0146342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>560</th>\n",
       "      <td>attributes</td>\n",
       "      <td>-0.013367</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>119</th>\n",
       "      <td>florianopolis</td>\n",
       "      <td>-0.0126672</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>525</th>\n",
       "      <td>disaggregated</td>\n",
       "      <td>-0.0117474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>580</th>\n",
       "      <td>gladstone</td>\n",
       "      <td>-0.0110999</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>428</th>\n",
       "      <td>recruitment</td>\n",
       "      <td>-0.0109947</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>328</th>\n",
       "      <td>bovine</td>\n",
       "      <td>-0.010766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>88</th>\n",
       "      <td>greenfield</td>\n",
       "      <td>-0.0106539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>331</th>\n",
       "      <td>pulsar</td>\n",
       "      <td>-0.0103355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>rumen</td>\n",
       "      <td>-0.0102316</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>394</th>\n",
       "      <td>pulsars</td>\n",
       "      <td>-0.00910533</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>153</th>\n",
       "      <td>ammonium</td>\n",
       "      <td>-0.00841409</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>305</th>\n",
       "      <td>aridity</td>\n",
       "      <td>-0.00702331</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>protozoa</td>\n",
       "      <td>-0.00663317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>baucau</td>\n",
       "      <td>-0.00659033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>234</th>\n",
       "      <td>visualization</td>\n",
       "      <td>-0.0062123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>485</th>\n",
       "      <td>nematodes</td>\n",
       "      <td>-0.0045522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>116</th>\n",
       "      <td>mosses</td>\n",
       "      <td>-0.00422636</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>597</th>\n",
       "      <td>waite</td>\n",
       "      <td>-0.00403283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>285</th>\n",
       "      <td>goethite</td>\n",
       "      <td>-0.00183042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>450</th>\n",
       "      <td>microbiology</td>\n",
       "      <td>-0.00113159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>avocado</td>\n",
       "      <td>-0.00110459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>288</th>\n",
       "      <td>consistency</td>\n",
       "      <td>-0.000649769</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>598 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           Keyword Similarity Score\n",
       "375          ocean              1.0\n",
       "535            sea         0.684473\n",
       "297         oceans         0.643771\n",
       "451          coast         0.577876\n",
       "360        coastal         0.544094\n",
       "186          waves         0.497618\n",
       "214          tidal         0.478499\n",
       "72           water         0.476902\n",
       "18           reefs         0.457699\n",
       "595       tropical          0.44709\n",
       "30          seabed         0.428529\n",
       "253           fish         0.397474\n",
       "415         marine         0.395392\n",
       "559           sand         0.395074\n",
       "416        weather         0.393229\n",
       "484           wind         0.382924\n",
       "533    continental         0.356296\n",
       "63       ecosystem         0.356034\n",
       "232        natural         0.353634\n",
       "474    temperature         0.351932\n",
       "395          atoll         0.348955\n",
       "309       salinity         0.346013\n",
       "356       sediment         0.342138\n",
       "12        wildlife         0.341011\n",
       "488        estuary         0.338248\n",
       "156           wave         0.331474\n",
       "34        location          0.33124\n",
       "208      sediments          0.32903\n",
       "306    environment         0.328656\n",
       "404       boundary         0.327967\n",
       "..             ...              ...\n",
       "28          barley       -0.0190345\n",
       "466          maize        -0.018247\n",
       "295         cereal       -0.0182227\n",
       "579          ferns       -0.0181958\n",
       "517       workflow       -0.0171294\n",
       "479       low-cost        -0.015994\n",
       "576   fluorescence       -0.0154273\n",
       "221      myriapods       -0.0146342\n",
       "560     attributes        -0.013367\n",
       "119  florianopolis       -0.0126672\n",
       "525  disaggregated       -0.0117474\n",
       "580      gladstone       -0.0110999\n",
       "428    recruitment       -0.0109947\n",
       "328         bovine        -0.010766\n",
       "88      greenfield       -0.0106539\n",
       "331         pulsar       -0.0103355\n",
       "47           rumen       -0.0102316\n",
       "394        pulsars      -0.00910533\n",
       "153       ammonium      -0.00841409\n",
       "305        aridity      -0.00702331\n",
       "105       protozoa      -0.00663317\n",
       "29          baucau      -0.00659033\n",
       "234  visualization       -0.0062123\n",
       "485      nematodes       -0.0045522\n",
       "116         mosses      -0.00422636\n",
       "597          waite      -0.00403283\n",
       "285       goethite      -0.00183042\n",
       "450   microbiology      -0.00113159\n",
       "106        avocado      -0.00110459\n",
       "288    consistency     -0.000649769\n",
       "\n",
       "[598 rows x 2 columns]"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Source Keyword: ocean' + '\\n')\n",
    "results_similarity = {}\n",
    "for keyword in unique_keywords:\n",
    "    try:\n",
    "        results_similarity[keyword] = str(model.similarity(keyword, 'ocean'))\n",
    "    except:\n",
    "        pass\n",
    "df = pd.DataFrame()\n",
    "df['Keyword'] = results_similarity.keys()\n",
    "df['Similarity Score'] = results_similarity.values()\n",
    "df.sort_values(by=['Similarity Score'], ascending=False)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## What else might we be able to do with word2vec vectors and keywords\n",
    "\n",
    "One thing that might be possible is to look at how closely related a particular set of keywords attributed to a dataset  actually are. We can compare word vector distance ourselves and so can calculate average distance of keywords with each other. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.29919001, -0.11731   , -0.0089925 , -0.37059   , -0.06722   ,\n",
       "        0.15163   , -0.061105  ,  0.29587001,  0.36515999, -1.50870001,\n",
       "        0.46160001, -0.15762   ,  0.015131  ,  0.31378999,  0.49033999,\n",
       "        0.23762   ,  0.27667001,  0.44819999, -0.64633   ,  0.66012001,\n",
       "       -0.65131003,  0.36984   , -0.41850001, -0.053622  , -0.0097837 ,\n",
       "       -0.12772   ,  0.47055   ,  0.65263999, -0.37119001,  0.48045999,\n",
       "        0.39286   , -0.06189   , -0.88924998, -0.55094999,  0.35034001,\n",
       "       -0.32709   ,  0.29975   , -0.056088  , -0.035726  ,  0.46678001,\n",
       "       -0.27544001, -0.01794   ,  0.41068   ,  0.16943   , -0.38510999,\n",
       "        0.29284   ,  0.51858997,  0.56309003,  0.24123   ,  0.099607  ,\n",
       "       -0.20422   ,  0.11269   , -0.49399999, -0.87515002, -0.31327   ,\n",
       "        0.43037   , -0.11785   ,  0.46608001,  0.13485   , -0.29508001,\n",
       "        0.071264  ,  0.31647   ,  1.10640001, -0.35238999,  0.11418   ,\n",
       "       -0.30375999, -0.56996   ,  0.70823997, -0.15449999,  0.16618   ,\n",
       "        0.1987    , -0.17903   , -0.19689   , -0.27891001, -0.76629001,\n",
       "        0.23562001,  0.69582999, -0.16906001,  0.57331002,  0.10815   ,\n",
       "       -0.19524001, -0.29041001, -0.59411001,  0.23573001,  0.20468   ,\n",
       "        0.47402999, -0.23048   , -0.15125   , -0.14287999, -0.56348002,\n",
       "       -0.10656   ,  0.40211999, -0.082997  , -0.22390001, -0.30127999,\n",
       "        0.67137998,  0.52389997,  0.37303999,  0.14026999,  0.16144   ,\n",
       "        0.43261999,  0.57077003,  0.42592999,  0.13321   , -0.44323999,\n",
       "        0.10222   ,  0.40268999,  0.27155   ,  0.14915   ,  0.024384  ,\n",
       "        0.071348  , -0.27318999, -0.053039  , -0.20828   ,  0.19783001,\n",
       "       -0.30609   ,  0.10706   ,  0.21874   ,  0.059144  ,  0.12594999,\n",
       "       -0.35730001, -0.83701998, -0.68713999,  0.0051108 ,  0.3795    ,\n",
       "        0.69436997,  0.095449  ,  0.03565   ,  0.52607   , -0.44367999,\n",
       "       -0.29462999,  0.23534   ,  0.25206   ,  0.54913002,  0.54696   ,\n",
       "        0.32997   ,  0.17747   , -0.010207  , -0.4082    , -0.056547  ,\n",
       "        0.24877   ,  0.1409    , -0.39412999, -0.31325001, -0.86926001,\n",
       "       -0.011931  ,  0.54620999,  0.51165003, -0.27309999, -0.0036448 ,\n",
       "        0.073265  ,  0.099931  ,  0.16576   , -0.37744999,  0.71210998,\n",
       "        0.60350001, -0.13561   , -0.40702   ,  0.16904999,  0.10413   ,\n",
       "        0.082192  , -0.15620001,  0.40612   ,  0.29192001,  0.010975  ,\n",
       "       -0.35427999,  0.13127001,  0.43439001, -0.073638  ,  0.75231999,\n",
       "       -0.078701  , -0.14765   ,  0.09442   ,  0.071877  , -0.19168   ,\n",
       "        0.052296  ,  0.047609  ,  0.24071001, -0.01385   , -0.36346999,\n",
       "       -0.12535   ,  0.13027   ,  0.26745   ,  0.23916   ,  0.57080001,\n",
       "       -0.76853001,  0.13985001,  0.35321   , -0.2098    , -0.16614   ,\n",
       "        0.12789001,  0.41624999, -0.032202  , -1.09630001,  0.26881   ,\n",
       "        0.50839001, -0.29831001, -1.10950005,  0.25488999, -0.22379   ,\n",
       "        1.26769996,  0.16649   , -0.19984999, -0.2811    ,  0.48842001,\n",
       "        0.36778   ,  0.18153   , -0.67558998,  0.30105001,  0.16327   ,\n",
       "        0.051021  , -0.25863001,  0.034524  , -0.38958001, -0.035517  ,\n",
       "       -0.1376    ,  1.01629996, -0.10668   , -0.32596999, -0.28246   ,\n",
       "        1.21920002,  0.41918001,  0.1242    ,  0.038983  , -0.21337999,\n",
       "       -0.34496   , -0.26998001, -0.73084998, -0.1814    , -0.51795   ,\n",
       "        0.52011001, -0.22322001, -0.22628   , -0.070672  ,  0.83191001,\n",
       "        0.19023   , -0.24928001, -0.1885    ,  0.021689  ,  0.31211999,\n",
       "       -0.70852   ,  0.23819   , -0.12503   ,  0.16992   , -0.75865   ,\n",
       "        0.016328  ,  0.11299   ,  0.0034248 , -0.032634  , -0.16235   ,\n",
       "       -0.58937001,  0.54838002,  0.40274   ,  0.23458   , -0.023596  ,\n",
       "        0.14429   ,  0.40711999,  0.13214999,  0.21811999,  0.50448   ,\n",
       "        0.068399  , -0.50648999, -0.012295  , -0.56260997, -0.2429    ,\n",
       "        0.19044   ,  0.071061  , -0.78464001, -0.31560001, -0.37182999,\n",
       "        0.31880999, -0.26536   ,  0.14229   ,  0.33750999, -0.084764  ,\n",
       "        0.54408997, -1.80040002, -0.020358  ,  0.21841   , -0.27943   ,\n",
       "       -0.46899   ,  0.66947001,  0.19723   , -0.46858999, -0.15442   ,\n",
       "       -0.62577999, -0.20599   , -0.081889  ,  0.36353999,  0.22763   ,\n",
       "       -0.91742003,  0.60233998, -0.078529  , -0.21348999, -0.36337   ,\n",
       "        0.63546997,  0.25263   , -0.22014   , -0.117     , -0.19186001], dtype=float32)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.word_vec('sea')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Average Similarity</th>\n",
       "      <th>Data Record Keywords</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>9.481299</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fcfa', 'keywords': ['soil', 'agriculture', 'crops', 'pastures', 'carbon', 'sequestration', 'baseline', 'stocks', 'fractions', 'nitrogen', 'carbonate', 'particulate', 'humus', 'resistant', 'charcoal']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.893878</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fd45', 'keywords': ['invertebrates', 'insects', 'biosecurity', 'taxonomy', 'arthropods', 'systematic']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8.847030</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fd04', 'keywords': ['temperature', 'salinity', 'depth', 'fluorescence', 'oxygen', 'PAR', 'ocean', 'phosphate', 'nitrate', 'ammonium', 'silicate', 'Australia, Western Australia coast', 'East Indian Ocean', 'Indian Ocean']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>8.826347</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c165387', 'keywords': ['malacology', 'molluscs', 'mollusc', 'chitons', 'clams', 'mussels', 'snails', 'nudibranchs', 'sea', 'slugs', 'tusk', 'shells', 'octopus', 'squid', 'fauna']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>8.782723</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fcc8', 'keywords': ['permanent plot', 'Wet Tropics', 'fire', 'soil', 'rainforest', 'cyclone', 'mortality', 'recruitment', 'growth rate']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>8.776973</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fd3f', 'keywords': ['taxonomy', 'systematic', 'evolution', 'phylogeny', 'nomenclature', 'biogeography', 'phylogeography', 'herbarium', 'tropical', 'tropics', 'identification keys']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>8.631260</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f922', 'keywords': ['Tasmania', 'plant', 'trait', 'functional', 'species', 'leaf', 'seed', 'height']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>8.587046</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f87d', 'keywords': ['pollen', 'anther', 'pollination', 'almond', 'prunus', 'honeybee', 'orchard', 'nuts', 'fruit', 'flower']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>8.569571</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fd35', 'keywords': ['parasites', 'parasitology', 'Acanthocephala', 'thorny-headed worms', 'Acari', 'ticks', 'mites', 'Anoplura', 'sucking lice', 'bacteria', 'Cestoda', 'tapeworms', 'Digenea', 'trematodes', 'Insecta', 'insects', 'Monogenea', 'trematodes', 'Hippoboscida', 'keds', 'Mallophaga', 'lice', 'Nematoda', 'roundworms', 'Pentastomida', 'tongue worms', 'Protozoa', 'Siphonaptera', 'fleas', 'collection', 'birds', 'reptiles', 'amphibians', 'mammals', 'fauna', 'wildlife', 'systematics', 'phylogenetics', 'taxonomy', 'biogeography', 'pathology', 'phylogeography']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>8.474214</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fd43', 'keywords': ['terrestrial vertebrate', 'birds', 'mammals', 'eggs', 'amphibians', 'reptiles', 'fauna', 'calls', 'wildlife', 'population genetics', 'biogeography', 'systematics', 'taxonomy', 'phylogenetics', 'genomics', 'phylogeography', 'evolutionary genetics']}]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Average Similarity  \\\n",
       "0            9.481299   \n",
       "1            8.893878   \n",
       "2            8.847030   \n",
       "3            8.826347   \n",
       "4            8.782723   \n",
       "5            8.776973   \n",
       "6            8.631260   \n",
       "7            8.587046   \n",
       "8            8.569571   \n",
       "9            8.474214   \n",
       "\n",
       "                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      Data Record Keywords  \n",
       "0                                                                                                                                                                                                                                                                                                                                                                                  [{'_id': '5915a6be3dd97b0a4c19fcfa', 'keywords': ['soil', 'agriculture', 'crops', 'pastures', 'carbon', 'sequestration', 'baseline', 'stocks', 'fractions', 'nitrogen', 'carbonate', 'particulate', 'humus', 'resistant', 'charcoal']}]  \n",
       "1                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   [{'_id': '5915a6be3dd97b0a4c19fd45', 'keywords': ['invertebrates', 'insects', 'biosecurity', 'taxonomy', 'arthropods', 'systematic']}]  \n",
       "2                                                                                                                                                                                                                                                                                                                                                             [{'_id': '5915a6be3dd97b0a4c19fd04', 'keywords': ['temperature', 'salinity', 'depth', 'fluorescence', 'oxygen', 'PAR', 'ocean', 'phosphate', 'nitrate', 'ammonium', 'silicate', 'Australia, Western Australia coast', 'East Indian Ocean', 'Indian Ocean']}]  \n",
       "3                                                                                                                                                                                                                                                                                                                                                                                                          [{'_id': '59159e9b3dd97b0a4c165387', 'keywords': ['malacology', 'molluscs', 'mollusc', 'chitons', 'clams', 'mussels', 'snails', 'nudibranchs', 'sea', 'slugs', 'tusk', 'shells', 'octopus', 'squid', 'fauna']}]  \n",
       "4                                                                                                                                                                                                                                                                                                                                                                                                                                                 [{'_id': '5915a6be3dd97b0a4c19fcc8', 'keywords': ['permanent plot', 'Wet Tropics', 'fire', 'soil', 'rainforest', 'cyclone', 'mortality', 'recruitment', 'growth rate']}]  \n",
       "5                                                                                                                                                                                                                                                                                                                                                                                                     [{'_id': '5915a6be3dd97b0a4c19fd3f', 'keywords': ['taxonomy', 'systematic', 'evolution', 'phylogeny', 'nomenclature', 'biogeography', 'phylogeography', 'herbarium', 'tropical', 'tropics', 'identification keys']}]  \n",
       "6                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     [{'_id': '5915a6bd3dd97b0a4c19f922', 'keywords': ['Tasmania', 'plant', 'trait', 'functional', 'species', 'leaf', 'seed', 'height']}]  \n",
       "7                                                                                                                                                                                                                                                                                                                                                                                                                                                             [{'_id': '5915a6bd3dd97b0a4c19f87d', 'keywords': ['pollen', 'anther', 'pollination', 'almond', 'prunus', 'honeybee', 'orchard', 'nuts', 'fruit', 'flower']}]  \n",
       "8  [{'_id': '5915a6be3dd97b0a4c19fd35', 'keywords': ['parasites', 'parasitology', 'Acanthocephala', 'thorny-headed worms', 'Acari', 'ticks', 'mites', 'Anoplura', 'sucking lice', 'bacteria', 'Cestoda', 'tapeworms', 'Digenea', 'trematodes', 'Insecta', 'insects', 'Monogenea', 'trematodes', 'Hippoboscida', 'keds', 'Mallophaga', 'lice', 'Nematoda', 'roundworms', 'Pentastomida', 'tongue worms', 'Protozoa', 'Siphonaptera', 'fleas', 'collection', 'birds', 'reptiles', 'amphibians', 'mammals', 'fauna', 'wildlife', 'systematics', 'phylogenetics', 'taxonomy', 'biogeography', 'pathology', 'phylogeography']}]  \n",
       "9                                                                                                                                                                                                                                                                                                              [{'_id': '5915a6be3dd97b0a4c19fd43', 'keywords': ['terrestrial vertebrate', 'birds', 'mammals', 'eggs', 'amphibians', 'reptiles', 'fauna', 'calls', 'wildlife', 'population genetics', 'biogeography', 'systematics', 'taxonomy', 'phylogenetics', 'genomics', 'phylogeography', 'evolutionary genetics']}]  "
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy, itertools, collections\n",
    "\n",
    "average_keyword_dist_record = {}\n",
    "\n",
    "def get_word_vec(word, word_vec):\n",
    "    a_vec = None\n",
    "    if word in word_vec.keys():\n",
    "        a_vec = word_vec[word]\n",
    "    else:\n",
    "        try:\n",
    "            a_vec = model.word_vec(word)\n",
    "        except:\n",
    "            a_vec = None\n",
    "        word_vec[word] = a_vec\n",
    "    return a_vec\n",
    "\n",
    "for record in keywords_dict:    \n",
    "    metadata_record_keywords = record['keywords']\n",
    "    total_distance = 0\n",
    "    count = 1\n",
    "    word_vec = {}\n",
    "    sets = [frozenset(pair) for pair in itertools.product(metadata_record_keywords, metadata_record_keywords)]\n",
    "    unique_pairs = set(sets)\n",
    "    for pair in unique_pairs:\n",
    "        pair = list(pair)\n",
    "        if len(pair) < 2:\n",
    "            continue\n",
    "        else:\n",
    "            a_vec = get_word_vec(pair[0], word_vec)\n",
    "            b_vec = get_word_vec(pair[1], word_vec)\n",
    "            if (a_vec is not None) and (b_vec is not None):\n",
    "                distance = numpy.linalg.norm(a_vec-b_vec)\n",
    "            else:\n",
    "                continue\n",
    "            count += 1\n",
    "            total_distance += distance\n",
    "    average_distance = total_distance / count\n",
    "    if average_distance in average_keyword_dist_record.keys():\n",
    "        average_keyword_dist_record[average_distance].append(record)\n",
    "    else:\n",
    "        average_keyword_dist_record[average_distance] = [record]\n",
    "\n",
    "result = collections.OrderedDict(sorted(average_keyword_dist_record.items(), reverse=True))\n",
    "df = pd.DataFrame()\n",
    "df[\"Average Similarity\"] = result.keys()\n",
    "df[\"Data Record Keywords\"] = result.values()\n",
    "df[0:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Average Similarity</th>\n",
       "      <th>Data Record Keywords</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>173</th>\n",
       "      <td>3.336291</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c16538b', 'keywords': ['fossils', 'fauna']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>174</th>\n",
       "      <td>3.336098</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f9f4', 'keywords': ['TERN_Soils', 'TERN_Soils_DSM', 'hematite', 'goethite', 'soil color', 'soil colour', 'visible-near-infrared reflectance', 'soil mapping', 'geostatistical simulations']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>175</th>\n",
       "      <td>3.204942</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f96a', 'keywords': ['geogenetics', 'Drupal', 'genomics', 'bioinformatics']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>176</th>\n",
       "      <td>3.151918</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c1653d8', 'keywords': ['reptiles', 'fauna']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>177</th>\n",
       "      <td>3.088527</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c165388', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653d5', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653de', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653f6', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c165401', 'keywords': ['fauna', 'mammals']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178</th>\n",
       "      <td>2.949659</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c1653bb', 'keywords': ['microbial', 'microbes']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>179</th>\n",
       "      <td>2.949578</td>\n",
       "      <td>[{'_id': '5915a6be3dd97b0a4c19fb3e', 'keywords': ['FGARA', 'maize', 'corn', 'AGRICULTURE-Crops', 'AGRICULTURE-Irrigation', 'Australia', 'INDUSTRY-Primary', 'LAND-Use', 'Queensland', 'SOIL', 'Flinders catchment', 'Gilbert catchment', 'land suitability']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>180</th>\n",
       "      <td>2.871381</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f968', 'keywords': ['AEM', 'airborne electromagnetic', 'Gascoyne', 'groundwater', 'aquifer', '']}]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>181</th>\n",
       "      <td>2.752079</td>\n",
       "      <td>[{'_id': '5915a6bd3dd97b0a4c19f930', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave height', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f931', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave rider buoy', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f932', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave height', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f935', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'Australian Sea Lion', 'Neophoca cinerea', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f936', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'New Zealand Fur Seal', 'Arctocephalus forsteri', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f939', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'marine benthic su...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>182</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>[{'_id': '59159e9b3dd97b0a4c165379', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c16537b', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c16537e', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c165380', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165381', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165382', 'keywords': ['microbes']}, {'_id': '59159e9b3dd97b0a4c16538c', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165391', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165395', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165399', 'keywords': ['Ferns']}, {'_id': '59159e9b3dd97b0a4c16539d', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a0', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c1653a1', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a2', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a4', 'keywords': ['microfungi', 'Bacteria', 'Viruses', 'microbes']}, {'_id': '59159e9b3d...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Average Similarity  \\\n",
       "173            3.336291   \n",
       "174            3.336098   \n",
       "175            3.204942   \n",
       "176            3.151918   \n",
       "177            3.088527   \n",
       "178            2.949659   \n",
       "179            2.949578   \n",
       "180            2.871381   \n",
       "181            2.752079   \n",
       "182            0.000000   \n",
       "\n",
       "                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        Data Record Keywords  \n",
       "173                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  [{'_id': '59159e9b3dd97b0a4c16538b', 'keywords': ['fossils', 'fauna']}]  \n",
       "174                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              [{'_id': '5915a6bd3dd97b0a4c19f9f4', 'keywords': ['TERN_Soils', 'TERN_Soils_DSM', 'hematite', 'goethite', 'soil color', 'soil colour', 'visible-near-infrared reflectance', 'soil mapping', 'geostatistical simulations']}]  \n",
       "175                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               [{'_id': '5915a6bd3dd97b0a4c19f96a', 'keywords': ['geogenetics', 'Drupal', 'genomics', 'bioinformatics']}]  \n",
       "176                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 [{'_id': '59159e9b3dd97b0a4c1653d8', 'keywords': ['reptiles', 'fauna']}]  \n",
       "177                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      [{'_id': '59159e9b3dd97b0a4c165388', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653d5', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653de', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c1653f6', 'keywords': ['mammals', 'fauna']}, {'_id': '59159e9b3dd97b0a4c165401', 'keywords': ['fauna', 'mammals']}]  \n",
       "178                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             [{'_id': '59159e9b3dd97b0a4c1653bb', 'keywords': ['microbial', 'microbes']}]  \n",
       "179                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           [{'_id': '5915a6be3dd97b0a4c19fb3e', 'keywords': ['FGARA', 'maize', 'corn', 'AGRICULTURE-Crops', 'AGRICULTURE-Irrigation', 'Australia', 'INDUSTRY-Primary', 'LAND-Use', 'Queensland', 'SOIL', 'Flinders catchment', 'Gilbert catchment', 'land suitability']}]  \n",
       "180                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         [{'_id': '5915a6bd3dd97b0a4c19f968', 'keywords': ['AEM', 'airborne electromagnetic', 'Gascoyne', 'groundwater', 'aquifer', '']}]  \n",
       "181  [{'_id': '5915a6bd3dd97b0a4c19f930', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave height', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f931', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave rider buoy', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f932', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'wave height', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f935', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'Australian Sea Lion', 'Neophoca cinerea', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f936', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'New Zealand Fur Seal', 'Arctocephalus forsteri', 'coast', 'coastal', 'Australia']}, {'_id': '5915a6bd3dd97b0a4c19f939', 'keywords': ['CAMRIS', 'Coastal and Marine Resources Information System', 'marine benthic su...  \n",
       "182  [{'_id': '59159e9b3dd97b0a4c165379', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c16537b', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c16537e', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c165380', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165381', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165382', 'keywords': ['microbes']}, {'_id': '59159e9b3dd97b0a4c16538c', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165391', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165395', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c165399', 'keywords': ['Ferns']}, {'_id': '59159e9b3dd97b0a4c16539d', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a0', 'keywords': ['microbial']}, {'_id': '59159e9b3dd97b0a4c1653a1', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a2', 'keywords': ['plants']}, {'_id': '59159e9b3dd97b0a4c1653a4', 'keywords': ['microfungi', 'Bacteria', 'Viruses', 'microbes']}, {'_id': '59159e9b3d...  "
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[-10:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Can we assess good keyword tagging this way? Could we score keyword descriptiveness as users enter them? \n",
    "\n",
    "(2.9496593475341797, [{'_id': '59159e9b3dd97b0a4c1653bb', 'keywords': ['microbial', 'microbes']}])\n",
    "\n",
    "vs \n",
    "\n",
    "(8.1640734042017922, [{'_id': '59159e9b3dd97b0a4c165439', 'keywords': ['herbarium', 'herbaria', 'plants', 'angiosperms', 'dicots', 'monocots', 'gymnosperms', 'pteridophytes', 'mosses', 'liverworts', 'lichens', 'fungi', 'algae', 'fossils', 'wood', 'microbes']}])\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualising clusters of keywords\n",
    "\n",
    "We can reduce the dimensions using principal component analysis (PCA) and provide a very (overly) simplified graph of relationships between words "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.datasets import fetch_20newsgroups\n",
    "import re\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Limiting the model\n",
    "Lets limit the vocabulary of the model to just keywords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "result_vocab = { key: val for key, val in model.vocab.items() if key in unique_keywords }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['world',\n",
       " 'economic',\n",
       " 'office',\n",
       " 'history',\n",
       " 'oil',\n",
       " 'water',\n",
       " 'building',\n",
       " 'fire',\n",
       " 'community']"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(result_vocab.keys())[1:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reducing the dimensions and plotting\n",
    "\n",
    "We can then reduce the dimensions from 300 to 2 and plot the relationship of keywords to other keywords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = model[result_vocab]\n",
    "\n",
    "font = matplotlib.font_manager.FontProperties(fname='./ipag.ttc')\n",
    "\n",
    "words_np = []\n",
    "#a list of labels (words)\n",
    "words_label = []\n",
    "for word, vector in result_vocab.items():\n",
    "    words_np.append(model[word])\n",
    "    words_label.append(word)\n",
    "        \n",
    "pca = PCA(n_components=2)\n",
    "tsne = TSNE(n_components=2, random_state=0)\n",
    "numpy.set_printoptions(suppress=True)\n",
    "Y = tsne.fit_transform(words_np)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams['figure.figsize'] = [10, 5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(Y[:, 0], Y[:, 1])\n",
    "for label, x, y in zip(result_vocab.keys(), Y[:, 0], Y[:, 1]):\n",
    "    plt.annotate(label, xy=(x, y), xytext=(0, 0), textcoords='offset points')\n",
    "plt.ylim(10, 0)\n",
    "plt.xlim(-30, 0)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Suggesting new keywords\n",
    "\n",
    "We could envisage a keyword tagging interface were we suggest new keywords based on the average distance from existing keywords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('dog', 0.921538770198822),\n",
       " ('cat', 0.9122815728187561),\n",
       " ('dogs', 0.7285658121109009),\n",
       " ('pet', 0.6637269258499146),\n",
       " ('cats', 0.6461218595504761),\n",
       " ('puppy', 0.5556367635726929),\n",
       " ('horse', 0.5517184138298035),\n",
       " ('animal', 0.531834065914154),\n",
       " ('pets', 0.5269761085510254),\n",
       " ('hound', 0.5094276666641235)]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector((model['dog'] + model['cat'])/2,topn=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you have tagged a data item with dog and cats maybe you also want to tag it with \"animal\" or \"pets\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('runoff', 0.8186976909637451),\n",
       " ('water', 0.8018253445625305),\n",
       " ('sewage', 0.5654644966125488),\n",
       " ('run-off', 0.546949565410614),\n",
       " ('groundwater', 0.5467579960823059),\n",
       " ('polluted', 0.5163611173629761),\n",
       " ('drainage', 0.515787661075592),\n",
       " ('irrigation', 0.5122992992401123),\n",
       " ('pollution', 0.5067722797393799),\n",
       " ('seawater', 0.4968295097351074),\n",
       " ('soil', 0.4866226017475128),\n",
       " ('drinking', 0.4837367534637451),\n",
       " ('waste', 0.4792274832725525),\n",
       " ('rainwater', 0.4654936194419861),\n",
       " ('moisture', 0.4654146730899811),\n",
       " ('reservoirs', 0.46503356099128723),\n",
       " ('rain', 0.46365755796432495),\n",
       " ('potable', 0.46122878789901733),\n",
       " ('rivers', 0.4577578902244568),\n",
       " ('reservoir', 0.45721039175987244),\n",
       " ('wastewater', 0.4544180631637573),\n",
       " ('waters', 0.4541703462600708),\n",
       " ('flooding', 0.44811367988586426),\n",
       " ('clean', 0.44566574692726135),\n",
       " ('rainfall', 0.43809616565704346),\n",
       " ('contaminated', 0.4357152581214905),\n",
       " ('flow', 0.4356193542480469),\n",
       " ('drain', 0.43446698784828186),\n",
       " ('contamination', 0.4323706030845642),\n",
       " ('sediment', 0.4322854280471802)]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector((model['water'] + model['runoff'])/2,topn=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('radiation', 0.8520972728729248),\n",
       " ('solar', 0.8336105346679688),\n",
       " ('ultraviolet', 0.6234352588653564),\n",
       " ('uv', 0.5582942962646484),\n",
       " ('infrared', 0.5512473583221436),\n",
       " ('thermal', 0.5436907410621643),\n",
       " ('cosmic', 0.5349714756011963),\n",
       " ('sunlight', 0.5338567495346069),\n",
       " ('emitted', 0.5294535756111145),\n",
       " ('electromagnetic', 0.5245305299758911)]"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector((model['solar'] + model['radiation'])/2,topn=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Limitations of word2vec\n",
    "\n",
    "There are apparently lots of relationships that aren't semantic but are more based on word proximity or spelling. Famously it is possible to use word2vec to find that the relation\"man is related to king\" applied to \"woman\" gives \"queen\"\n",
    "It doesn't work so well with everything\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('king', 0.8065859079360962),\n",
       " ('queen', 0.6896163821220398),\n",
       " ('monarch', 0.5575490593910217),\n",
       " ('throne', 0.5565374493598938),\n",
       " ('princess', 0.5518684387207031),\n",
       " ('mother', 0.5142154097557068),\n",
       " ('daughter', 0.513315737247467),\n",
       " ('kingdom', 0.5025345087051392),\n",
       " ('prince', 0.5017740726470947),\n",
       " ('elizabeth', 0.4908031225204468),\n",
       " ('wife', 0.4840559959411621),\n",
       " ('crown', 0.4728339910507202),\n",
       " ('woman', 0.4675373435020447),\n",
       " ('her', 0.45041656494140625),\n",
       " ('royal', 0.4489114582538605),\n",
       " ('marry', 0.4381888508796692),\n",
       " ('married', 0.43084365129470825),\n",
       " ('sister', 0.42904388904571533),\n",
       " ('husband', 0.42386510968208313),\n",
       " ('ii', 0.4199441075325012),\n",
       " ('mary', 0.41993027925491333),\n",
       " ('duke', 0.41978520154953003),\n",
       " ('anne', 0.4154040217399597),\n",
       " ('regent', 0.41426482796669006),\n",
       " ('niece', 0.41178426146507263),\n",
       " ('marriage', 0.40905052423477173),\n",
       " ('margaret', 0.4079943299293518),\n",
       " ('she', 0.4075208306312561),\n",
       " ('emperor', 0.4063114523887634),\n",
       " ('duchess', 0.40345293283462524),\n",
       " ('widow', 0.4019556939601898),\n",
       " ('herself', 0.4018828868865967),\n",
       " ('reign', 0.3991032540798187),\n",
       " ('pregnant', 0.39799222350120544),\n",
       " ('isabella', 0.3978602886199951),\n",
       " ('mistress', 0.39719516038894653),\n",
       " ('granddaughter', 0.39549678564071655),\n",
       " ('consort', 0.3951156735420227),\n",
       " ('kings', 0.3949219584465027),\n",
       " ('grandmother', 0.3942447304725647),\n",
       " ('bhumibol', 0.3932991623878479),\n",
       " ('iii', 0.39300358295440674),\n",
       " ('ruler', 0.3924878239631653),\n",
       " ('eldest', 0.3922141194343567),\n",
       " ('crowned', 0.3866226077079773),\n",
       " ('vi', 0.38607141375541687),\n",
       " ('daughters', 0.3823390305042267),\n",
       " ('aragon', 0.38215020298957825),\n",
       " ('coronation', 0.38017725944519043),\n",
       " ('son', 0.37840408086776733),\n",
       " ('vii', 0.3768041729927063),\n",
       " ('female', 0.37576228380203247),\n",
       " ('cousin', 0.3757569193840027),\n",
       " ('birth', 0.37179118394851685),\n",
       " ('majesty', 0.3717663288116455),\n",
       " ('coretta', 0.3700453042984009),\n",
       " ('haakon', 0.3689405918121338),\n",
       " ('heir', 0.36295783519744873),\n",
       " ('women', 0.3623914122581482),\n",
       " ('marrying', 0.3622637093067169),\n",
       " ('lady', 0.3622066378593445),\n",
       " ('kalākaua', 0.3614727854728699),\n",
       " ('countess', 0.36107009649276733),\n",
       " ('empress', 0.3609313666820526),\n",
       " ('jordan', 0.3608926832675934),\n",
       " ('child', 0.35840916633605957),\n",
       " ('father', 0.3577882647514343),\n",
       " ('frederick', 0.3577301800251007),\n",
       " ('monarchy', 0.3573784828186035),\n",
       " ('anna', 0.3556566536426544),\n",
       " ('prussia', 0.3546048402786255),\n",
       " ('monarchs', 0.3541788160800934),\n",
       " ('marie', 0.3541252017021179),\n",
       " ('grandson', 0.35402578115463257),\n",
       " ('adulyadej', 0.35201650857925415),\n",
       " ('dynasty', 0.351993203163147),\n",
       " ('viii', 0.35128194093704224),\n",
       " ('navarre', 0.35096338391304016),\n",
       " ('diana', 0.34967494010925293),\n",
       " ('sultan', 0.3439084589481354),\n",
       " ('iv', 0.34327131509780884),\n",
       " ('rania', 0.3417149782180786),\n",
       " ('accompanied', 0.33802369236946106),\n",
       " ('kamehameha', 0.33767250180244446),\n",
       " ('luther', 0.3375445008277893),\n",
       " ('henry', 0.3373536467552185),\n",
       " ('eleanor', 0.33728939294815063),\n",
       " ('wedding', 0.3372892141342163),\n",
       " ('girl', 0.3360154628753662),\n",
       " ('widowed', 0.33549001812934875),\n",
       " ('nobles', 0.3342302441596985),\n",
       " ('edward', 0.333000123500824),\n",
       " ('catherine', 0.33172351121902466),\n",
       " ('rulers', 0.3316349685192108),\n",
       " ('pharaoh', 0.3310440182685852),\n",
       " ('alfonso', 0.33089131116867065),\n",
       " ('louise', 0.330410361289978),\n",
       " ('parents', 0.3303326368331909),\n",
       " ('goddess', 0.3299383521080017),\n",
       " ('palace', 0.3296911120414734)]"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector((model['woman'] + model['king'] - model['man']), topn=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('zucchini', 0.5701828002929688),\n",
       " ('eggplant', 0.5665943026542664),\n",
       " ('tomatoes', 0.5604736804962158),\n",
       " ('avocado', 0.5497550964355469),\n",
       " ('spinach', 0.5222511291503906),\n",
       " ('cucumber', 0.5068976283073425),\n",
       " ('lettuce', 0.4866516590118408),\n",
       " ('watermelon', 0.4849747121334076),\n",
       " ('onion', 0.48183757066726685),\n",
       " ('vegetable', 0.4807186722755432)]"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['yellow', 'tomato'], negative=['red'])\n",
    "#model.similar_by_vector(((model['yellow'] + model['strawberry'] - model['red']) + (model['yellow'] + model['tomato'] - model['red']))/2 , topn=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('water', 0.49908646941185),\n",
       " ('sanitation', 0.392291396856308),\n",
       " ('electricity', 0.3892342448234558),\n",
       " ('potable', 0.3729451894760132),\n",
       " ('sewage', 0.34720665216445923),\n",
       " ('emergency', 0.332465797662735),\n",
       " ('flood', 0.32895636558532715),\n",
       " ('drinking', 0.30943164229393005),\n",
       " ('irrigation', 0.3018989562988281),\n",
       " ('sewerage', 0.295976459980011),\n",
       " ('waters', 0.2916158437728882),\n",
       " ('supply', 0.29060104489326477),\n",
       " ('heaters', 0.28887104988098145),\n",
       " ('power', 0.288494348526001),\n",
       " ('forth', 0.2883579134941101),\n",
       " ('floodwaters', 0.28483453392982483),\n",
       " ('toilets', 0.2824137210845947),\n",
       " ('reservoir', 0.2806611657142639),\n",
       " ('shelter', 0.2799522578716278),\n",
       " ('inflatable', 0.2797348201274872),\n",
       " ('hose', 0.2758522629737854),\n",
       " ('fuel', 0.2753504514694214),\n",
       " ('hoses', 0.2748646140098572),\n",
       " ('sewer', 0.27269724011421204),\n",
       " ('supplies', 0.27109020948410034),\n",
       " ('boats', 0.26907333731651306),\n",
       " ('wastewater', 0.268878310918808),\n",
       " ('dams', 0.26764512062072754),\n",
       " ('dam', 0.26609787344932556),\n",
       " ('generators', 0.26593878865242004),\n",
       " ('electrical', 0.2656600773334503),\n",
       " ('stricken', 0.2646414041519165),\n",
       " ('yangtze', 0.26297667622566223),\n",
       " ('tankless', 0.26259127259254456),\n",
       " ('flow', 0.2618275582790375),\n",
       " ('conservation', 0.26107943058013916),\n",
       " ('generator', 0.2605353593826294),\n",
       " ('energy', 0.2603209912776947),\n",
       " ('rescue', 0.2601647973060608),\n",
       " ('grid', 0.26014894247055054),\n",
       " ('pylons', 0.25891825556755066),\n",
       " ('pollution', 0.25860434770584106),\n",
       " ('harness', 0.25708523392677307),\n",
       " ('floodwater', 0.2568952441215515),\n",
       " ('ocean', 0.25627079606056213),\n",
       " ('protection', 0.2557043433189392),\n",
       " ('groundwater', 0.255459189414978),\n",
       " ('wind', 0.25532370805740356),\n",
       " ('rooftops', 0.25519925355911255),\n",
       " ('boat', 0.25518953800201416),\n",
       " ('mains', 0.25508958101272583),\n",
       " ('feet', 0.2544407248497009),\n",
       " ('pump', 0.25398895144462585),\n",
       " ('grids', 0.25390052795410156),\n",
       " ('rooftop', 0.25368818640708923),\n",
       " ('hyacinth', 0.2529239058494568),\n",
       " ('cauvery', 0.25091350078582764),\n",
       " ('pressurized', 0.25051236152648926),\n",
       " ('leaping', 0.2501595914363861),\n",
       " ('infrastructure', 0.2501058578491211),\n",
       " ('pumps', 0.2500118315219879),\n",
       " ('debris', 0.24917924404144287),\n",
       " ('mosquito', 0.24900537729263306),\n",
       " ('control', 0.24862414598464966),\n",
       " ('vessel', 0.24789880216121674),\n",
       " ('rivers', 0.24772389233112335),\n",
       " ('reservoirs', 0.24760836362838745),\n",
       " ('waterworks', 0.24731683731079102),\n",
       " ('cooling', 0.2469479888677597),\n",
       " ('garbage', 0.24643045663833618),\n",
       " ('poor', 0.2464272677898407),\n",
       " ('across', 0.2462703138589859),\n",
       " ('stationary', 0.2442305088043213),\n",
       " ('barge', 0.24416176974773407),\n",
       " ('creature', 0.24321338534355164),\n",
       " ('areas', 0.24275937676429749),\n",
       " ('providing', 0.24173066020011902),\n",
       " ('ballast', 0.24130673706531525),\n",
       " ('drainage', 0.24097877740859985),\n",
       " ('lighting', 0.24094180762767792),\n",
       " ('leaky', 0.23956118524074554),\n",
       " ('substation', 0.2394288331270218),\n",
       " ('containment', 0.23862630128860474),\n",
       " ('flooding', 0.23751398921012878),\n",
       " ('pipes', 0.23732790350914001),\n",
       " ('edf', 0.23695628345012665),\n",
       " ('shelters', 0.23664331436157227),\n",
       " ('feeding', 0.23513272404670715),\n",
       " ('firefighting', 0.23498941957950592),\n",
       " ('utility', 0.23488032817840576),\n",
       " ('renewable', 0.23483392596244812),\n",
       " ('environment', 0.23482711613178253),\n",
       " ('waterfall', 0.23445864021778107),\n",
       " ('rainwater', 0.2337319254875183),\n",
       " ('knocking', 0.23371027410030365),\n",
       " ('flooded', 0.2334759533405304),\n",
       " ('tanks', 0.23306500911712646),\n",
       " ('rudimentary', 0.23276090621948242),\n",
       " ('sail', 0.2326579988002777),\n",
       " ('cooler', 0.23263385891914368)]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector((model['water'] - model['salt']), topn=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tomato\n",
      "red\n",
      "sky\n",
      "blue\n",
      "coal\n",
      "black\n",
      "sulfur\n",
      "yellow\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('sulfur', 0.4082890748977661),\n",
       " ('biomass', 0.37060636281967163),\n",
       " ('methane', 0.3680614233016968),\n",
       " ('liquefaction', 0.36776357889175415),\n",
       " ('coal', 0.3645560145378113),\n",
       " ('revivifying', 0.3642549216747284),\n",
       " ('einsteinium', 0.35991495847702026),\n",
       " ('furnaces', 0.35524335503578186),\n",
       " ('dioxide', 0.3489069938659668),\n",
       " ('so2', 0.3484277129173279),\n",
       " ('feedstock', 0.3460592031478882),\n",
       " ('workover', 0.34195971488952637),\n",
       " ('nitrogen', 0.3403649926185608),\n",
       " ('2,615', 0.3397219479084015),\n",
       " ('rooper', 0.3366228938102722),\n",
       " ('pollutant', 0.3364943861961365),\n",
       " ('kaltim', 0.3359118700027466),\n",
       " ('vrbovec', 0.3354317843914032),\n",
       " ('by-product', 0.33457881212234497),\n",
       " ('coking', 0.33435937762260437),\n",
       " ('co2', 0.33394262194633484),\n",
       " ('over-population', 0.3335029184818268),\n",
       " ('smelters', 0.3328727185726166),\n",
       " ('llw', 0.3318973183631897),\n",
       " ('geosmin', 0.3318478465080261),\n",
       " ('inti', 0.33148398995399475),\n",
       " ('diammonium', 0.3310140371322632),\n",
       " ('petrochemistry', 0.32994911074638367),\n",
       " ('nonylphenol', 0.32862165570259094),\n",
       " ('anthropogenic', 0.32846957445144653),\n",
       " ('adgas', 0.3280569016933441),\n",
       " ('stiassny', 0.3280256688594818),\n",
       " ('walerian', 0.3261450529098511),\n",
       " ('minggong', 0.32583990693092346),\n",
       " ('benta', 0.3252028822898865),\n",
       " ('bermanbraun', 0.3243739604949951),\n",
       " ('positronium', 0.3243127763271332),\n",
       " ('syngas', 0.3234598636627197),\n",
       " ('egr', 0.32331421971321106),\n",
       " ('mortola', 0.32194358110427856),\n",
       " ('geothermal', 0.32145506143569946),\n",
       " ('incinerators', 0.3207094073295593),\n",
       " ('engles', 0.320182204246521),\n",
       " ('pollutants', 0.3196779489517212),\n",
       " ('pimpri', 0.31882214546203613),\n",
       " ('ironworking', 0.3181834816932678),\n",
       " ('poultries', 0.3177013695240021),\n",
       " ('ogma', 0.3172787129878998),\n",
       " ('energy-rich', 0.3172415494918823),\n",
       " ('butcheries', 0.31687235832214355),\n",
       " ('riffage', 0.3168190121650696),\n",
       " ('toxicants', 0.3163517415523529),\n",
       " ('hcfc-22', 0.3163313567638397),\n",
       " ('iron-making', 0.3155806064605713),\n",
       " ('landfills', 0.3155641257762909),\n",
       " ('35p', 0.31463485956192017),\n",
       " ('salaspils', 0.31458011269569397),\n",
       " ('cybersource', 0.31361785531044006),\n",
       " ('shivling', 0.31358474493026733),\n",
       " ('melzner', 0.31334006786346436),\n",
       " ('297.6', 0.3133262097835541),\n",
       " ('yanzhou', 0.31332170963287354),\n",
       " ('ponnary', 0.31301116943359375),\n",
       " ('ethanol', 0.312915563583374),\n",
       " ('18:47', 0.31275713443756104),\n",
       " ('sportacus', 0.3127308189868927),\n",
       " ('3,786', 0.31262674927711487),\n",
       " ('barite', 0.3125482499599457),\n",
       " ('gosman', 0.31203407049179077),\n",
       " ('smelter', 0.31190574169158936),\n",
       " ('rollbacks', 0.3118346035480499),\n",
       " ('9:19', 0.311764657497406),\n",
       " ('emissions', 0.311617910861969),\n",
       " ('cerebrally', 0.31153470277786255),\n",
       " ('over-consumption', 0.31148260831832886),\n",
       " ('gelugor', 0.3114244341850281),\n",
       " ('paulescu', 0.3110874891281128),\n",
       " ('ekong', 0.3108476400375366),\n",
       " ('tellurium', 0.31065380573272705),\n",
       " ('gases', 0.31059587001800537),\n",
       " ('tongchuan', 0.31030917167663574),\n",
       " ('combustion', 0.3100438416004181),\n",
       " ('diefenderfer', 0.30998003482818604),\n",
       " ('ringforts', 0.30996185541152954),\n",
       " ('hechuan', 0.30900537967681885),\n",
       " ('fluorination', 0.30882591009140015),\n",
       " ('dibutyl', 0.30877870321273804),\n",
       " ('cfsci', 0.3087501525878906),\n",
       " ('byproducts', 0.3083183765411377),\n",
       " ('3,689', 0.30795249342918396),\n",
       " ('harue', 0.30777591466903687),\n",
       " ('vernie', 0.3077542781829834),\n",
       " ('giygas', 0.30749768018722534),\n",
       " ('geochemistry', 0.3071141242980957),\n",
       " ('liquefied', 0.30709195137023926),\n",
       " ('smelting', 0.3069117069244385),\n",
       " ('dyckerhoff', 0.30685168504714966),\n",
       " ('valdimarsson', 0.3068090081214905),\n",
       " ('chiren', 0.3067239820957184),\n",
       " ('greenhouse', 0.30667349696159363)]"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "colors = { 'tomato': 'red',\n",
    "          'sky' : 'blue',\n",
    "          'coal' : 'black',\n",
    "          'sulfur' : 'yellow'         \n",
    "         }\n",
    "difference = None\n",
    "for word1, word2 in colors.items():\n",
    "    print(word1)\n",
    "    print(word2)\n",
    "    if difference is None:\n",
    "        difference = (model[word1] - model[word2])\n",
    "    else:\n",
    "        difference += (model[word1] - model[word2])\n",
    "\n",
    "difference /= len(colors)\n",
    "\n",
    "model.similar_by_vector(difference,topn=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('fruits', 0.6204125285148621),\n",
       " ('flowers', 0.5955206155776978),\n",
       " ('bright', 0.5717793703079224),\n",
       " ('red', 0.570356011390686),\n",
       " ('berries', 0.5619615912437439),\n",
       " ('purple', 0.550870418548584),\n",
       " ('vegetables', 0.5429374575614929),\n",
       " ('orange', 0.5417463779449463),\n",
       " ('pink', 0.5412436723709106),\n",
       " ('green', 0.5293484926223755),\n",
       " ('flower', 0.5223943591117859),\n",
       " ('tomato', 0.5201548337936401),\n",
       " ('colored', 0.5142461061477661),\n",
       " ('juice', 0.5093603134155273),\n",
       " ('dried', 0.5090316534042358),\n",
       " ('apples', 0.5070767998695374),\n",
       " ('ripe', 0.49489259719848633),\n",
       " ('tomatoes', 0.49187809228897095),\n",
       " ('trees', 0.4913737177848816),\n",
       " ('beans', 0.4900389313697815),\n",
       " ('edible', 0.48981401324272156),\n",
       " ('colors', 0.4891403317451477),\n",
       " ('grapes', 0.48298153281211853),\n",
       " ('leaves', 0.4816040098667145),\n",
       " ('nuts', 0.48078590631484985),\n",
       " ('mushrooms', 0.47849369049072266),\n",
       " ('grape', 0.4782850444316864),\n",
       " ('blue', 0.47625142335891724),\n",
       " ('stuff', 0.47623932361602783),\n",
       " ('picking', 0.47455766797065735),\n",
       " ('varieties', 0.4745110273361206),\n",
       " ('cherry', 0.47390830516815186),\n",
       " ('olive', 0.4728095829486847),\n",
       " ('lots', 0.47081437706947327),\n",
       " ('vegetable', 0.47075340151786804),\n",
       " ('sweet', 0.46875250339508057),\n",
       " ('olives', 0.4683453440666199),\n",
       " ('herbs', 0.4678265452384949),\n",
       " ('leaf', 0.46765732765197754),\n",
       " ('dark', 0.4664687514305115),\n",
       " ('carrots', 0.4657733142375946),\n",
       " ('fresh', 0.4654398560523987),\n",
       " ('honey', 0.4647597670555115),\n",
       " ('seeds', 0.46344590187072754),\n",
       " ('eat', 0.46244674921035767),\n",
       " ('bananas', 0.4621644616127014),\n",
       " ('wild', 0.46022021770477295),\n",
       " ('white', 0.46021589636802673),\n",
       " ('usually', 0.45945098996162415),\n",
       " ('onions', 0.4590492844581604),\n",
       " ('bunches', 0.4585489332675934),\n",
       " ('chopped', 0.458248496055603),\n",
       " ('soft', 0.4544278383255005),\n",
       " ('whole', 0.45395177602767944),\n",
       " ('potatoes', 0.4535807967185974),\n",
       " ('colorful', 0.45232927799224854),\n",
       " ('roses', 0.4506530463695526),\n",
       " ('taste', 0.4494629502296448),\n",
       " ('candy', 0.44932830333709717),\n",
       " ('color', 0.4492695927619934),\n",
       " ('blossoms', 0.44774413108825684),\n",
       " ('eaten', 0.4468792676925659),\n",
       " ('oranges', 0.44654470682144165),\n",
       " ('lemon', 0.4461049437522888),\n",
       " ('like', 0.445486843585968),\n",
       " ('strawberries', 0.4447556436061859),\n",
       " ('tree', 0.4437260031700134),\n",
       " ('fragrant', 0.4432797431945801),\n",
       " ('citrus', 0.4431683123111725),\n",
       " ('cream', 0.44218528270721436),\n",
       " ('banana', 0.44126826524734497),\n",
       " ('flesh', 0.4411826431751251),\n",
       " ('black', 0.44096997380256653),\n",
       " ('cheese', 0.43874573707580566),\n",
       " ('eating', 0.43809062242507935),\n",
       " ('foliage', 0.4376506209373474),\n",
       " ('ones', 0.43680763244628906),\n",
       " ('onion', 0.4367557764053345),\n",
       " ('cherries', 0.4363901615142822),\n",
       " ('pretty', 0.435855507850647),\n",
       " ('small', 0.431800901889801),\n",
       " ('watermelon', 0.4313318133354187),\n",
       " ('eggs', 0.43113943934440613),\n",
       " ('peaches', 0.4307435154914856),\n",
       " ('shiny', 0.4289301335811615),\n",
       " ('carrot', 0.42849868535995483),\n",
       " ('creamy', 0.4276190996170044),\n",
       " ('mango', 0.4275202751159668),\n",
       " ('you', 0.427245169878006),\n",
       " ('stuffed', 0.42686811089515686),\n",
       " ('petals', 0.42657914757728577),\n",
       " ('sometimes', 0.42646563053131104),\n",
       " ('guys', 0.425609827041626),\n",
       " ('grown', 0.4247550666332245),\n",
       " ('salad', 0.42435842752456665),\n",
       " ('bread', 0.4236631691455841),\n",
       " ('flavor', 0.42288124561309814),\n",
       " ('potato', 0.4220251142978668),\n",
       " ('peppers', 0.4212028384208679),\n",
       " ('dry', 0.4199395179748535)]"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.most_similar(positive=['yellow', 'bunch', 'fruit'], topn=100) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('meters', 0.8691878318786621),\n",
       " ('tall', 0.8533526062965393),\n",
       " ('metres', 0.7684553265571594),\n",
       " ('feet', 0.7459859848022461),\n",
       " ('meter', 0.7151063680648804),\n",
       " ('height', 0.6292724013328552),\n",
       " ('metre', 0.6133023500442505),\n",
       " ('m', 0.6051520705223083),\n",
       " ('inches', 0.5688554048538208),\n",
       " ('centimeters', 0.5521728992462158),\n",
       " ('ft', 0.5430169105529785),\n",
       " ('foot', 0.5379891395568848),\n",
       " ('taller', 0.5375359058380127),\n",
       " ('centimetres', 0.5244764089584351),\n",
       " ('thick', 0.5189315676689148),\n",
       " ('diameter', 0.5047868490219116),\n",
       " ('cm', 0.5028142333030701),\n",
       " ('towering', 0.49395322799682617),\n",
       " ('above', 0.47632113099098206),\n",
       " ('tallest', 0.4680122137069702),\n",
       " ('6-foot', 0.46595123410224915),\n",
       " ('yards', 0.4580432176589966),\n",
       " ('stands', 0.4558824896812439),\n",
       " ('kilometers', 0.45539915561676025),\n",
       " ('10-foot', 0.4520391821861267),\n",
       " ('200', 0.44777730107307434),\n",
       " ('trees', 0.4474383592605591),\n",
       " ('jump', 0.44673773646354675),\n",
       " ('cubic', 0.4390564560890198),\n",
       " ('standing', 0.4371226131916046),\n",
       " ('20-foot', 0.4334980249404907),\n",
       " ('wide', 0.43289780616760254),\n",
       " ('length', 0.43161168694496155),\n",
       " ('tower', 0.4310564398765564),\n",
       " ('150', 0.43039873242378235),\n",
       " ('weighs', 0.4296438694000244),\n",
       " ('kilometer', 0.4279867708683014),\n",
       " ('miles', 0.4276895225048065),\n",
       " ('100', 0.42602118849754333),\n",
       " ('kilograms', 0.424282968044281),\n",
       " ('distance', 0.42420727014541626),\n",
       " ('erect', 0.4224482476711273),\n",
       " ('kilometres', 0.42055001854896545),\n",
       " ('5-foot', 0.42014312744140625),\n",
       " ('elevation', 0.41819339990615845),\n",
       " ('mile', 0.41770315170288086),\n",
       " ('30-foot', 0.4128264784812927),\n",
       " ('width', 0.41116052865982056),\n",
       " ('walls', 0.4099467694759369),\n",
       " ('beneath', 0.4079047441482544),\n",
       " ('50', 0.406992107629776),\n",
       " ('km', 0.40332144498825073),\n",
       " ('square', 0.40266430377960205),\n",
       " ('1,500', 0.4026317298412323),\n",
       " ('110', 0.40239906311035156),\n",
       " ('kilometre', 0.40236175060272217),\n",
       " ('inch', 0.4023545980453491),\n",
       " ('8-foot', 0.4019384980201721),\n",
       " ('60', 0.40141525864601135),\n",
       " ('altitude', 0.40138956904411316),\n",
       " ('vertical', 0.4011596739292145),\n",
       " ('400', 0.399842232465744),\n",
       " ('surrounded', 0.39776724576950073),\n",
       " ('3-foot', 0.39742571115493774),\n",
       " ('fence', 0.3971388041973114),\n",
       " ('15-foot', 0.3947412371635437),\n",
       " ('4-foot', 0.3941405713558197),\n",
       " ('long', 0.3932751417160034),\n",
       " ('centimeter', 0.3927207887172699),\n",
       " ('around', 0.390997052192688),\n",
       " ('1,000', 0.39054566621780396),\n",
       " ('12-foot', 0.38922905921936035),\n",
       " ('bronze', 0.38897985219955444),\n",
       " ('towers', 0.3872654438018799),\n",
       " ('roof', 0.3824279010295868),\n",
       " ('measuring', 0.381745308637619),\n",
       " ('pounds', 0.3815544545650482),\n",
       " ('40', 0.3812980651855469),\n",
       " ('floors', 0.38095545768737793),\n",
       " ('40-foot', 0.3808055818080902),\n",
       " ('300', 0.3792579174041748),\n",
       " ('lanky', 0.3790246248245239),\n",
       " ('storeys', 0.3783241808414459),\n",
       " ('upright', 0.3775135278701782),\n",
       " ('800', 0.3756723403930664),\n",
       " ('distances', 0.3752571940422058),\n",
       " ('atop', 0.37509286403656006),\n",
       " ('ft.', 0.37490367889404297),\n",
       " ('130', 0.3739216923713684),\n",
       " ('120', 0.3734913468360901),\n",
       " ('concrete', 0.37315088510513306),\n",
       " ('rectangular', 0.37303727865219116),\n",
       " ('1.5', 0.37212395668029785),\n",
       " ('180', 0.37137654423713684),\n",
       " ('water', 0.3709507882595062),\n",
       " ('frame', 0.37079888582229614),\n",
       " ('seconds', 0.37059465050697327),\n",
       " ('220', 0.36907342076301575),\n",
       " ('wooden', 0.3689190745353699),\n",
       " ('nearly', 0.36841917037963867)]"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.similar_by_vector(model['meters'] + model['tall'], topn=100) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('fire', 0.5652563571929932),\n",
       " ('cease', 0.390556275844574),\n",
       " ('firing', 0.3865264058113098),\n",
       " ('gunfire', 0.3802269399166107),\n",
       " ('fires', 0.3728276789188385),\n",
       " ('gunmen', 0.357028603553772),\n",
       " ('fired', 0.3469422459602356),\n",
       " ('destroyed', 0.3426634967327118),\n",
       " ('explosions', 0.3339660167694092),\n",
       " ('civilians', 0.333923876285553)]"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# what is the direction to move in to be the opposite of hot?\n",
    "# a fire without its heat must move in the direction of \"cold\"\n",
    "# \n",
    "model.similar_by_vector(model['fire'] - model['hot'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}