feat(YouTube - Keyword filter): Add syntax to match whole keywords and not substrings

[LisoUseInAIKyrios]

Adds syntax to match keywords by word boundaries.
This syntax now allows short phrases like ai that previously could not be used.

How to use:

Surround any keyword or phrase with double quotes, such as:

Keyword: "ai"

Example videos that will be hidden:

• Is AI bad?
• Are you getting on AI's hype train?
• New #ai1337 model release!

But unrelated videos will not be hidden:

• Guide to DMCA fair use
• What air quality means to you
   

Keyword: "vice news"

Example videos that will be hidden:

• The bankruptcy of Vice News

But unrelated videos will not be hidden:

• Today's Advice Newscast

 
If you have keywords in languages that do not use spaces between words (Chinese, Japanese, Burmese, etc), then using whole word syntax is not recommended. But for all other languages it is recommended whenever possible since the filtering is much more accurate.

Using quotes will interfere with most pluralized words, as "fox" will no longer match foxes, so multiple variations may be needed for some words. But punctuation and other non letters are ignored, so "fox" will match and hide fox's, fox?, and #fox123.

Integration changes

[oSumAtrIX]

The user may not intuitively understand this behavior. The description should mention that only entire words are matched. Perhaps it is possible to add a toggle to control whether to match words or also substrings

[LisoUseInAIKyrios]

Yes the description should mention that both singular and plural keywords can be added.

I thought about adding a toggle to enable/disable whole word matching. But if toggled off then it's back to the old 3 character minimum length, and the under 3 character toast or the settings description would be a lot more complicated. Or remove the character limit even when whole word matching is off, but that will likely hide way too much stuff without the user realizing it.

I think requiring plurals might be the simplest, and a side effect of that is always using the more accurate keyword matching.

[LisoUseInAIKyrios]

I realized that whole word matching will act weird or fail when used with Chinese and other pictograph languages, since spacing around words is not common.

So whole word matching should be an on/off toggle.

[KobeW50]

Maybe whole word matching can be set with an operator, such as %, similar to how the $ and ^ operators were used in #2682

This will allow users to specify which words should only be filtered by whole-word matching and which should be filtered through the general matching scheme. This solves the issue with pictographic languages, and makes it so users don't need to add the plural version of words for words that they don't use whole-word matching on.

Example:

%ai (will only filter "ai" and not a word such as "rain")
robot (will filter "robot" and "robots")

[LisoUseInAIKyrios]

@KobeW50 Yeah that sounds interesting.

Would need to use syntax that nobody would ever want as part of a keyword itself (or would need to support escaping the syntax).

How to apply to a multiword phrase and not a single word, like ai show where you don't want to hide unrelated stuff like Thai showbiz?
 

Edit: The more complex quoting scheme below was not pursued, and a simple outer quote only was used.

Maybe use quotes as that is easy to understand and it implies a literal meaning. It also would allow matching each side of a multi-word phrase:

"ai" hides only ai.
It does not match any substring of another word.

"ai show" hides only ai show.
It does not match fail show, ai showbiz, or anything else.

"ai" show hides ai show, ai shows, ai showbiz, etc.
But does not match fail show.

ai "show" hides fail show, Thai show, etc.
But not match ai showbiz.

Using quotes that are not part of the start/end of a phrase would be ignored since it effectively would do nothing.
 

Or just keep it really simple and only support quotes around the entire word/phrase. That might be the simplest and maybe all that's needed. Only allowing quotes around the entire word/phrase also removes the issue of allowing quotes for whatever reason, because then "I love "air quotes"" would allow matching the exact string of I love "air quotes"

[LisoUseInAIKyrios]

I'll give the quotes idea a go and see how that works.

[oSumAtrIX]

You can employ escaping in order to allow the user to use quotes as the string to filter.

Escaping " with \ could simply be treated as part of the string to filter.
In ReVanced CLI dev branch I have written a parser (see the test package) to convert strings into any primitive type and lists of these types. There i used " to convert integers as literal strings for example but using \ i can escape the " and count that as part of the value.

[LisoUseInAIKyrios]

I will try using quotes around an entire word/phrase, as I think the selective whole word matching is too complex and not needed.

If quotes on the outside are the only valid place for syntax, then there is no need for escaping because plain quotes anywhere else are left as-is and treated as part of the keyword/phrase.