wikimedia / drafttopic

Predicting topics to new drafts based on Wikiprojects on English Wikipedia
https://drafttopic.readthedocs.io
MIT License
10 stars 7 forks source link

Wikidata drafttopic #54

Closed dibyaaaaax closed 4 years ago

dibyaaaaax commented 4 years ago

T254289

halfak commented 4 years ago

This is looking really good. I think it just needs a model build. You should be able to build the model on ores-misc-01.

halfak commented 4 years ago

Looks like the build is failing because you're not pulling in the wikidata embeddings when before the tests run. See https://github.com/wikimedia/drafttopic/blob/master/.travis.yml#L23 You'll need a line there for "make"ing the wikidata embeddings.

dibyaaaaax commented 4 years ago

Looks like the build is failing because you're not pulling in the wikidata embeddings when before the tests run. See https://github.com/wikimedia/drafttopic/blob/master/.travis.yml#L23 You'll need a line there for "make"ing the wikidata embeddings.

I previously assumed that the build was failing because the embeddings didn't exist in the server. After adding a line to make the embeddings I suppose that's the issue now?

halfak commented 4 years ago

Right. Looks like we need to build the embeddings and get them uploaded to the right spot for datasets.wikimedia.org.

If you tell me where to find the embeddings you built, I can help get them in the right place.

dibyaaaaax commented 4 years ago

@halfak it was built on ores-misc-01. you can also find it here.

halfak commented 4 years ago

OK I moved the files. They should be available within 30 minutes. Once that happens, we'll want to restart the travis build and make sure it all comes together. Then we'll be able to merge. Thanks!

dibyaaaaax commented 4 years ago

thanks for the help with the files @halfak :)

halfak commented 4 years ago

Nice work. @chrisalbon FYI, with the merging of this, there is now a wikidata-based topic model available for deployment in ORES. Let me know if you want to talk about what it would take to get @dibyaaaaax's work into production.