Xavi, I am impressed that you could in Softcatalà get enough bilingual texts to create an English-Catalan neural translator. Congratulations on the results! I am curious to know how big the corpus you collected has been, as well as from which sources to ensure the quality of the translations. I'd maybe add that probably it would not be possible to collect such a corpus for Valencian Catalan, so I guess we face in this neural translator a typical problem with lesser-user languages/varieties. If it is ever considered necessary to generate Valencian, this will have to be done by translating it into "reference" Catalan and then automatically adapting it. In fact the same happens for the many flavours we currently have in Apertium for Catalan, both Valencian and "Catalonian". By the way, is Softcatalà trying to create a neural translator for the Spanish-Catalan pair? Hèctor
Missatge de Xavi Ivars <xavi.iv...@gmail.com> del dia dg., 18 d’oct. 2020 a les 0:04: > I actually think apertium eng-cat may be one of the english-romanic pairs > that has received more care. Marc has worked quite a lot on it since he did > it during a GSoC, but has continued working on his it adding newly > developer modules (perceptron tagger, separable, lexical selection, > anaphora resolution,...) > > I agree with Mikel that the evaluation is well made. And after having > tried the neural a few times, it will probably be hard to get there, in > terms if fluency, with Apertium. Translations with Apertium are, usually, > more "robotic". > > -- > Xavi Ivars > < http://xavi.ivars.me > > > El dv., 16 d’oct. 2020, 19:35, Tanmai Khanna <khanna.tan...@gmail.com> va > escriure: > >> Not to get too defensive haha but in the talk makes certain statements, >> such as "Neural machine translation is more fluent and adequate compared to >> RBMT" (not verbatim), and later on a comparison of post-editability, but >> doesn't really comment on the status of these tools, i.e. how much data >> were the NMT systems trained on, and how much is the English-Catalan system >> of Apertium is worked on (which we know is not a lot), so I was just >> surprised to see general statements against RBMT. Not saying this as a >> member of Apertium but generally evaluations like this should be done >> carefully and all the factors should be considered before comparing >> multiple systems, without that, it's easy to arrive at false conclusions. >> >> My two cents :) >> *तन्मय खन्ना * >> *Tanmai Khanna* >> >> >> On Thu, Oct 15, 2020 at 9:27 PM Mikel L. Forcada <m...@dlsi.ua.es> wrote: >> >>> Dear Apertiumers: >>> >>> here's a 20-minute talk from Vicent Briva where he evaluates Apertium >>> English–Catalan in comparison with the SoftCatalà neural engine and >>> Google Translate. >>> >>> I think the evaluation is quite well made. >>> >>> https://www.youtube.com/watch?v=IiRVhAYpecw >>> >>> We do not fare too well but, hey, we know this language pair needs love. >>> >>> All the best, >>> >>> >>> Mikel >>> >>> >>> >>> _______________________________________________ >>> Apertium-stuff mailing list >>> Apertium-stuff@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/apertium-stuff >>> >> _______________________________________________ >> Apertium-stuff mailing list >> Apertium-stuff@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/apertium-stuff >> > _______________________________________________ > Apertium-stuff mailing list > Apertium-stuff@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/apertium-stuff >
_______________________________________________ Apertium-stuff mailing list Apertium-stuff@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/apertium-stuff
