Michael created this task. Michael added projects: Wikidata, revscoring. Restricted Application added a subscriber: Aklapper. Restricted Application added projects: Scoring-platform-team, artificial-intelligence.
TASK DESCRIPTION In the process of training ORES/articlequality AI to estimate the quality of Wikidata items, we are using a set of features <https://github.com/wikimedia/articlequality/blob/master/articlequality/feature_lists/wikidatawiki.py> including some that are depending on the `datatype` of a statement. However, it turns out that the `datatype` is currently never available (`None`) in the actual data that we are working with. This means that some of the features are just plain not working and our model isn't as good as it could be. Extensive digging revealed that the problem is that we are using the "wrong" API endpoint. Currently, we are using `action=query&prop=revisions&rvprop=content` which gives us the plain content of the revision as it is stored in the database. However, we do not store a properties `datatype` in the serialization of an item in the database and thus it is not available in the request-response. **acceptance criteria:** [ ] The property datatype in statements is available for feature analysis **Notes:** - That probably has to be implemented by changing which api is used by the revscoring Extractor <https://github.com/wikimedia/revscoring/blob/a3b22bb9641cd370f37d6ce798f9f3a1cdd1925d/revscoring/extractors/api/extractor.py>, maybe to a native Wikibase API? - it should be considered whether to use some of the existing caching - but also it needs to be considered how these changes affect the dumps-based analysis (which might use same API) - but also the dumps based analysis should maybe use entities dump instead of xml dump TASK DETAIL https://phabricator.wikimedia.org/T259351 EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/ To: Michael Cc: Lydia_Pintscher, Ladsgroup, Aklapper, Michael, Akuckartz, darthmon_wmde, jeropbrenda, Nandana, Lahi, Gq86, Xinbenlv, Vacio, Capankajsmilyo, GoranSMilovanovic, Fz-29, QZanden, LawExplorer, _jensen, rosalieper, Scott_WUaS, notconfusing, Wikidata-bugs, aude, Ricordisamoa, Alchimista, He7d3r, Mbch331
_______________________________________________ Wikidata-bugs mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs
