From: Sergio O. R. <ser...@gm...> - 2007-12-21 08:47:28
|
Hola Quiz=E1s lo que necesitemos sea una forma de combinar paradigmas que no =20= implique un producto cartesiano. Sea t una transducci=F3n. Lo que ocurre ahora mismo es que si se =20 concatenan dos o m=E1s paradigmas P y Q ocurre lo siguiente: P =3D p1,p2... pP Q =3D q1,q2... qQ PQ =3D p1q1, p1q2, ... p1qQ,p2q1 ... pPqQ Sean P' y Q' dos "paradigmas especiales" (el nombre que le ponga lo =20 vais a tumbar) que cumplan la restricci=F3n de que |P'| =3D |Q'|. Se =20= define la concatenaci=F3n de estos dos paradigmas especiales as=ED: P'Q' =3D p'1q'1, p'2q'2, ..., p'kq'k Esto se generaliza f=E1cilmente para el caso de transducciones =20 intermedias: P'tQ' =3D p1tq1, p2tq2, ..., pktqk etc. Esto no necesita de pretaggers ni de nada parecido. Saludos El 21/12/2007, a las 6:54, Mikel L. Forcada escribi=F3: > Hi there, > Have read the thing. The only thing that I worry about is whether it =20= > is possible > to generate the de-mangling dictionary used before generation easily =20= > from a > regular "mangled" dictionary. What worries me is moving labels =20 > across paradigm > calls (would need recursive processing). > > Would that work when morphemes come in different order for different =20= > forms of a > word? In basque, the present form for "I see it" is "dakusat" where =20= > the "d-" > stands for "it" and the "-t" for "I", but "I saw it" is "nekusan" =20 > where "n-" > stands for "I" and there is no overt representation for "it". > > Mikel > > ---- Missatge original ---- > >> I had an idea this evening about how we might hack around dealing =20 >> with >> prefix inflection. I tried it out and put the results here. Would >> welcome any comments. >> >> = http://xixona.dlsi.ua.es/wiki/index.php/Partial_hack_for_prefix_inflection= >> >> Fran >> >> >> = ------------------------------------------------------------------------- >> This SF.net email is sponsored by: Microsoft >> Defy all challenges. Microsoft(R) Visual Studio 2005. >> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ >> _______________________________________________ >> Apertium-stuff mailing list >> Ape...@li... >> https://lists.sourceforge.net/lists/listinfo/apertium-stuff >> > > Mikel L. Forcada <ml...@dl...> > Dept. Llenguatges i Sistemes Inform=E0tics > Universitat d\'Alacant, E-03071 Alacant (Spain) > Tel.: +34 96 590 9776 Fax: +34 96 590 9326 > > = ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2005. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Apertium-stuff mailing list > Ape...@li... > https://lists.sourceforge.net/lists/listinfo/apertium-stuff |