El dj 27 de 01 de 2011 a les 13:03 +0000, en/na Jimmy O'Regan va escriure: > On 27 January 2011 09:07, Hèctor Alòs i Font <[email protected]> wrote: > > 2011/1/27 Francis Tyers <[email protected]> > >> > >> > >> I've added "Windows" as a proper noun, "do so" is difficult as I think > >> it should translate as "hacerlo", but then the pronoun moves if it is > >> finite, e.g. "lo hará", added "more than half of" as a determiner (as > >> with "a shitload of") > >> > >> 1) More than half of software developers are already building > >> applications for Windows 7 and nearly 80% will do so within the > >> next year, a new survey has found. > >> > >> web) Más que medio de software developers ya está construyendo > >> aplicaciones para Ventanas 7 y nearly 80% hará tan dentro del año > >> próximo, una encuesta nueva ha encontrado. > >> > >> svn) Más de la mitad de desarrolladores de software ya están > >> construyendo aplicaciones para Windows 7 y casi 80% hará tan dentro > >> del el año que viene, una encuesta nueva ha encontrado. > >> > > > > > > As you know, proper names are a mess, so we have to store big amounts of > > them to avoid being ridiculous. Unfortunately I couldn't yet catch a list of > > trade marks, as Windows, > > Every country has a registry for trademarks, so it should at least be > possible to get lists of them. > > > but enterprises names (from the Fortune top 500 > > list), first names, family names and place names (the lasts from > > French-speaking and Catalan-speaking countries). So Pierre Noël better not > > be translated into e.g. Stone Christmas :-) (by the way, I still have lots > > of problems with the French "Marie" analyzed as a verb, while fortunately > > "Paris" already is not got as a common name) > > > > You may find the lists I got at > > https://apertium.svn.sourceforge.net/svnroot/apertium/incubator/apertium-eo-fr/lexic/ > > https://apertium.svn.sourceforge.net/svnroot/apertium/trunk/apertium-eo-ca/lexic/ > > (but I couldn't them yet fully process) > > > > (For enterprises I got the top 500 list of Fortune, but e.g. in > > http://fr.transnationale.org/ is said to be 7,000 names - I don't know if > > they can be got somehow) > > > > It should be better for all of us if the could put somewhere together > > lexical resources in order to facilitate their finding and reuse in the > > translators (maybe a directory on sourceforge exits for that, but I don't > > know). I would prefer to have plain files with a clear explanation of their > > content rather than Apertium-formatted stuff because every translator > > ITYM *not* every > > > prefers to differenciate first names from family names or not, masculine > > first names from feminine ones or not, maybe someone would want to add the > > 5,000 most often Catalan family names (from Idescat) for his/her X-Y > > translator or not, and so on. From plane text files, anyone can quickly > > generate Apertium-like files. Lexical resources should be also referenced in > > the Apertium wiki. > > > > By the way, does someone of us yet used Wikipedia as a resource for huge > > amounts of proper nouns (with possibly translations into other languages)? > > > > Antonio Toral presented a paper at FreeRBMT about exactly that. > > > In connection with proper noun translation, we have a problem in Apertium > > when dealing with regular expressions. If I include, for instance, this > > couple of regular expressions (in just one dictionary): > > <e> <re>Saint\-[A-Z][a-z]+</re><i></i><par n="Andorre__np"/></e> > > <e> <re>Sainte\-[A-Z][a-z]+</re><i></i><par n="Andorre__np"/></e> > > the compilation (lt-comp) slows down from 12 s. to 34 s. in my computer. If > > the expression is more complicated (in order to deal with non-English > > characters) it's even worst. I simply can't use them in most of the cases. > > It also slows down processing and increases memory usage, as was > discussed here before.
Is it something that can be improved / fixed ? Fran ------------------------------------------------------------------------------ Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)! Finally, a world-class log management solution at an even better price-free! Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY! http://p.sf.net/sfu/arcsight-sfd2d _______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
