Great, Tanmai! Probably, it would be better to generate the pronoun in t2x rather that t1x, I think, but the new module apertium-recursive ( http://wiki.apertium.org/wiki/Apertium-recursive ) should be taken into account too. I couldn't try it yet, but it looks very powerful. By the way, one of the problems we meet is that we have to be sure that the subject is not yet in the sentence (for instance, "el hombre que vino ayer" = "the man who came yesterday" but "el hombre que escucharon ayer" = "the man they listen yesterday").
Missatge de Tanmai Khanna <[email protected]> del dia dl., 4 de nov. 2019 a les 21:54: > Hey Hèctor, > I'm in the process of implementing conditional anaphora resolution where > you can define multiple anaphors. We can define verbs as a type of anaphor > and give the rules the appropriate scores to guess a reference(subject in > this case). > > This should work decently because Anaphora resolution is to a large extent > trying to guess the subject. > > After guessing a subject we just have to add the appropriate preposition. > I'm not sure if this would be done in t1x or t2x. I remember thinking about > this and I think we decided t2x is the place to do this, and if that's the > case we need to find a way to pass ref to t2x also. Or if we can find a way > to add a subject in t1x that would be way easier. > > Let me know what you think. > > Tanmai > > Get Outlook for Android <https://aka.ms/ghei36> > > ------------------------------ > *From:* Hèctor Alòs i Font <[email protected]> > *Sent:* Monday, November 4, 2019 11:44:49 PM > *To:* [apertium-stuff] <[email protected]> > *Subject:* [Apertium-stuff] Null-subject resolution > > Thanks to the past GSoC, we have the new Anaphora-resolution module, what > is great. Nevertheless, it does not cover the problem of getting the > reference of the subject of a verb in null-subject languages. We have quite > a lot of problems when translating from a null-subject language (e.g. > Catalan, Occitan, Spanish, Italian) into a language with explicit subject > (e.g. French or English), because we can only guess between he/she/it, > il(s)/elle(s), etc. Could it be possible to think about a tool dealing with > this problem for the next GSoC? > Hèctor > _______________________________________________ > Apertium-stuff mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/apertium-stuff >
_______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
