Ksarasola added a comment.

Some ideas I could share with other participants in the Hackaton:

Some experiments to create an MT system to translate captions of images in Commons from Spanish to Basque and from Basque to Spanish, could be adapted to other language pairs.
1. Collecting all the images with  captions in both Basque and  Spanish available in Wikimedia Commons. (done for Spanish-Basque)
2. Adapting a general purpose NMT system to the domain of  image captions..  (done for Spanish-Basque)
3. Creating a service to help in the translation of captions using the NMT service.  (to be done, I need help)
4. Semi-automatic tagging of each photo with one of those 11 different kinds of image (Person, HumanGroup, Place/Location, Institution, Building, AnimalPlant, Event/sport, History, Map/Icon, Culture, and Others). I think that this information could be useful to increase translation quality. As Common's categories are not reliable, iwe are  we extracting this information from Wikipedia and Wikidata. (in development)

If successful, this experiments could create new tools to help to supply Commons contents in more languages.

TASK DETAIL
https://phabricator.wikimedia.org/T191025

EMAIL PREFERENCES
https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: Keegan, Ksarasola
Cc: Ksarasola, ESM, Mholloway, Abbe98, Aklapper, matthiasmullie, Cparle, MarkTraceur, Abit, SandraF_WMF, Lahi, PDrouin-WMF, Gq86, E1presidente, Ramsey-WMF, GoranSMilovanovic, QZanden, Tramullas, Acer, LawExplorer, Culex, Susannaanas, Aschroet, Jane023, Wikidata-bugs, PKM, Base, aude, Ricordisamoa, Lydia_Pintscher, Fabrice_Florin, Raymond, Steinsplitter, Mbch331
_______________________________________________
Wikidata-bugs mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs

Reply via email to