A 2015-12-22 19:12, Joan escrigué: > Francis,and all > > I've recently found out that pairs can be enabled without a real > "release". FANTASTIC!! > > Can you give me a list of what improvements you expect to see ? Is it > just the vocabulary? YES. BESIDES THE FILE _DOCUMENT.__TXT_ THAT > I ATTACH YOU, THERE ARE SOME WORDS AS BIOGRAPHIE, FILMOGRAPHIE, > NEVEU, PETIT-FILS... I didn't understand what you meant by "strange > characters". THEY ARE MAINLY CREATED WHEN ORIGINALLY IT FIND "<REF > NAME="XX">. THE “STRANGE CHARACTERS ARE: >SPAN, ABBR CLAS, > CONTENDITABLE...YOU CAN SEE FOR EXEMPLE IN > HTTPS://CA.WIKIPEDIA.ORG/W/INDEX.PHP?TITLE=BOB_RAFELSON&ACTION=EDIT > [9] VERSUS > HTTPS://FR.WIKIPEDIA.ORG/W/INDEX.PHP?TITLE=BOB_RAFELSON&ACTION=EDIT > [10]
This is a problem for the ContentTranslation team. It seems like the problem has to do with templates. > Would anyone be against me converting fr-ca to: > > * three-letter codes > * monolingual language packages > * adding lexical selection support > On our part, NO PROBLEM > ? Ok. Here are the problems I find: * Using the large French lexicon from either br-fr or fr-es causes a substantial decrease in translation quality. * The tagger is really bad. Here is a comparison of before/after. http://paste2.org/9GhW8LKe I can fix the testvoc errors, it will probably take around 2-3 days. I can also expand the lexicon from crossing fr-es and es-ca. However, I will need help with: 1) Fixing the tagging. You can write constraint grammar rules, or you can manually annotate texts. If you manually annotate texts, we will need around 1500 sentences annotated to make a substantial improvement 2) Proofreading the dictionary. 3) Writing lexical selection rules. If you would be interested in working on this, I think we can get something releasable in 2-3 full days of work. Let me know if you would be interested and we can pick the days to meet up on IRC. Fran ------------------------------------------------------------------------------ _______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
