Sergio Ortiz Rojas <[email protected]> čálii: > Hello, > > I’m going to complete a bit the variant functionality of lt-comp. > > 1. We want to have the possibility to mark an entry with more than one > variant simultaneously. In XML, the way to go is to define the > attribute as NMTOKEN, and define variants as identifiers, > space-separated in the v attribute. So where now you need to do: > > <e v=“valencian”><i>escombra</i><par n=“casa”/></e> > <e v=“balearic”><i>escombra</i><par n=“casa”/></e> > <e><i>granera</i><par n=“casa/></e> > > After the modification is done, you will be able to do as well: > > <e v=“valencian balearic”><i>escombra</i><par n=“casa”/></e> > <e><i>granera</i><par n=“casa/></e> > > So you can compile lt-comp -v (-vr, -vl) whatever variant and it should work. > > 2. And other minor change. As you know, we tend to accept (LR) > everything from every variant, but generate (RL) just one variant, the > one we specify. But some variants create problems (i.e. Balearic > variant generates a lot of additional homography in standard and > valencian Catalan, and this problem has caused its exclusion > historically in Apertium), so I want to add a --exclude-variant option > in lt-comp to forbid explicitly variants that we don’t want even to > analyze. The exclusion will be nondestructive, so if the entry belongs > to more than one variant and at least one of them is not excluded, it > will be included at least as analysis (LR). > > Any suggestion/piece of advice?
I like these changes :-) One question: If you have <e v="balearic"><i>xyz</i><par n="casa"/></e> where the form xyz *only* exists in balearic (and thus doesn't create homography), would it still be removed? OTOH, maybe you wouldn't mark such an entry with a v="" …
signature.asc
Description: PGP signature
------------------------------------------------------------------------------ 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
