Page MenuHomePhabricator

Investigate potential of using "database reports" to generate suggestions
Closed, ResolvedPublic

Description

This task involves the work of investigating the extent to which the Editing Team could use database reports as a source for generating "compelling and constructive" suggestions (T404220.

Decision(s) to be made

  • What – if any – aspects of database reports could the Editing Team leverage in the creation of edit suggestions?

Learning prompts

A non-exhaustive list...

  • How are the lists contained with database reports generated?
  • How many wikis seems to be leveraging database reports generated?
  • What – if any – reports align with the types of suggestions we are seeking to offer within Suggestion Mode?
  • What types of editing actions do the lists lend themselves to (e.g. adding sources, formatting, expanding articles, etc.)?
  • Approximately how many suggestions are contained with these database reports?
  • What format are these reports made available within? Might there be a programmatic way we could transform the logic used to build these reports into suggestions we can surface within VE's Suggestion Mode?
  • How accurate are the entries in practice — do they genuinely represent actionable edit opportunities?

Requirements

  • Using the "Learning prompts" above as a guide, arrive to (and document) a point of view on the === Decision(s) to be made

Event Timeline

Here's what I found.

How it works: each report you see on that index page is using its own instance of the {{Database report}} template, with a custom SQL query. A bot runs the queries (of any properly formatted template invocation) on a specified basis against a replica of the wikipedia database and updates the tables.

It seems like its value is in that it's an easy way for anyone to execute queries against a replica DB and get nice, consistent report tables. So I don’t know if this template itself is useful to us; the main purpose of the bot/template/module is the creation of the reports themselves. If we were to actually use some of this, the usefulness would end at executing the SQL queries.

That being said, the numbers we see here could inform what we choose to look into for suggestions. I found about 15-20 reports that could lead to relevant suggestions, which you can see in detail on my notes.

I found about 15-20 reports that could lead to relevant suggestions, which you can see in detail on my notes.

Wonderful! Thank you, @medelius. I've added the ideas you linked to above to the Edit Check/Suggestion Ideas spreadsheet.