Page MenuHomePhabricator

Improve TextMatchEditCheck
Closed, ResolvedPublicBUG REPORT

Description

TextMatchEditCheck is an existing experimental check that we've been using for testing / demos, mostly as an example of a non-AI pseudo tone check ("unfortunately").

It lets you configure rules in its config section in MediaWiki:editcheck-config.json for what to search for, which it internally calls replacers. Despite the name, the only action it allows you to take is to delete the matched content.

Stories

As a technical contributor to Wikipedia motivated to improve the content of Wikipedia, I would would value making people who are editing aware of the text deletions and replacements they could consider making in an easeful and structured way so that I can scale the impact I can have on improving the quality of Wikipedia's content.

Requirements

MVP

  • Provide more actions that can be taken upon a match
  • Allow replacers to specify their own visibility config, which would be any of the generic config supported by BaseEditCheck (e.g. account status, edit count, category, template on page, namespace, specific page-name, etc.)
  • Enhance the way "dismiss" works so that if you dismiss one specific instance of a suggestion within a specific range (e.g. replace "color" with "colour" in the lead section of the Double-decker bus article) it does not dismiss other instances of other text suggestion matches within the same range

Original list

  • Provide more actions that can be taken upon a match
  • Make replacer pattern-matching more sophisticated by giving users a way to provide regular expressions as well as just literal strings.
  • Allow replacers to specify their own visibility config, which would be any of the generic config supported by BaseEditCheck (e.g. account status, edit count, category, template on page, namespace, specific page-name, etc.)
  • Also, implement some of those visibility config settings mentioned above
  • Enhance the way "dismiss" works so that dismissing one replacer won't dismiss all textmatch checks
  • Let users specify the severity for a specific replacer rule?

References

  • Notes where @medelius is documenting the current status of TextMatchEditCheck.

Event Timeline

Change #1191446 had a related patch set uploaded (by Medelius; author: Medelius):

[mediawiki/extensions/VisualEditor@master] TextMatchEditCheck visibility configs

https://gerrit.wikimedia.org/r/1191446

Change #1191446 merged by jenkins-bot:

[mediawiki/extensions/VisualEditor@master] TextMatchEditCheck visibility configs

https://gerrit.wikimedia.org/r/1191446

Now that visibility configs have been implemented in 1191446, could we focus on implementing the replacement action capability next? This would:

  • Directly support the "common typos" use case mentioned by the Telugu community. They shared the following feedback

We would like a tool to evaluate article quality against known standards and a local list of common typos.

Maybe there's something to be learned from this tool: https://replacer.toolforge.org/stats (available for Spanish and Galician Wikipedias). It shows common grammatical errors or typos and allows you to replace them or edit them if an editor considers it is required

  • Build on the existing WIP replacement patch referenced
  • Provide more value than the current delete-only functionality

@medelius @DLynch - What are your thoughts on prioritizing replacements next? Are there any technical considerations or specifications that we should be aware of before proceeding?

Thanks for the helpful context. I'm looking into the replacements now - I agree it's a next logical step. I'll follow up if there are any gaps I come across.

Thank you for your prompt response @medelius ! Happy to learn you see value in this <3

Change #1193268 had a related patch set uploaded (by Medelius; author: Medelius):

[mediawiki/extensions/VisualEditor@master] TextMatchEditCheck improvements

https://gerrit.wikimedia.org/r/1193268

Next step(s)

Per today's team planning meeting, the next steps are as follows:

  • @medelius to to remove default configurations from 1193268, merge, and deploy via backport alongside T404470
  • @medelius to file ticket and begin work on "make case sensitivity i18n-safe"

Change #1193268 merged by jenkins-bot:

[mediawiki/extensions/VisualEditor@master] TextMatchEditCheck improvements

https://gerrit.wikimedia.org/r/1193268

medelius updated the task description. (Show Details)
DLynch added a project: Skipped QA.

There's follow-ups, but this is done. It has also been tested as part of other tasks.