Xavi Ivars <[email protected]> wrote: > [-- multipart/alternative, Encoding 7bit, 61 Zeilen --] > > [-- text/plain, Encoding quoted-printable, Zeichensatz: UTF-8, 28 Zeilen > --] > > Missatge de Flammie <[email protected]> del dia dg., 5 d’ag. 2018 a les 14:54: > >> >> [...] >> >> There is a -signed weight system in e.g. openfst that was added exactly >> for this intuition of weights, so it operates on R-, min, - instead of >> R+, max, +, maybe it could be an ok solution? >> >> > You mean still having "bigger weights, worse" but using negative numbers > (which, in this case would be that "smaller abs(x)" would be worse (as > they're actually bigger), but then we would be consistent with lexical > selection?
Pretty much yeah, this is what I found most logical of the options given by the WFSTs when talking about this interpretation before. > That would work, I guess. In any case, as long as there's a reason to do it > the way it is now, I'm OK. It was more an "out of curiosity" question, and > some bias towards consistency that I have, than a real concern. Consistency is definitely an important issue, not the least as I just came from intensive two weeks of teaching this stuffs. I'm not sure which is the best solution though, another possibility would be not to re-use the @w attribute with different interpretation but to create a new @p (penalty) for the concept. > We just need to make sure that this is properly documented. Exactly, and for a GSOC project, we should probably have documentation as a part of final things to do, I believe there is some already in the wiki even? -- Doktor Tommi A Pirinen, Computational Linguist, <https://flammie.github.io/purplemonkeydishwasher/>, Universität Hamburg, Hamburger Zentrum für Sprachkorpora <http://hzsk.de>. CLARIN-D Entwickler. President of ACL SIGUR SIG for Uralic languages <http://gtweb.uit.no/sigur/>. I tend to follow inline-posting style in desktop e-mail messages. ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
