Per Tunedal <[email protected]> writes: > Does that mean that the monolingual dictionaries now are independent of > the language pairs? What about the old requirement that all words in the > monodix had to be translated for the pair; i.e. words had to be present > in both monodixies and in the bidix. Is that requirement now abandoned? > What happens when translating to Swedish if a form in the foreign > language is missing in Swedish of vice versa?
The analyser is automatically trimmed to the bidix on compilation, so if a word is in swe.dix but not swe-dan.dix/swe-dan.autobil.bin, then swe-dan.automorf.bin will not have that word. However, we don't take the generator into account when trimming, so you can still get the #-type generation errors, e.g. if a word is in swe.dix and swe-dan.dix but not dan.dix. However, some times these #-errors are from bad transfer rules. Since the tagsets can differ, and it's up to transfer to ensure consistency, I don't think there's any sane way to auto-trim the bidix+analyser to the generator … > Is it now possible to extend the Swedish dictionary, without having to > extend the Danish dictionary at the same time? If so, it would > facilitate contributions considerably. Lars Aronsson would be happy. That is possible now, yes. However, if you change *existing* entries, you can mess up other pairs, so there are some whole new possible problems :-) (For example, if you decide that "hennes" really should be a <det><poss> instead of <prn><gen>, but the pair swe-foo still has <prn><gen> in the bidix.) >> (The diff for the Danish side is 67736 lines long, so that may be more >> of a challenge to merge … but I'd still say it's worth it to merge the >> Swedish side right away.) >> > > The next step after I merged the sv-da.sv.dix with the swe-dix would be > to merge with the pair is-sv. In that way both the pair sv-da and the > pair is-sv would benefit from corrections in the Swedish monodix. Yes :-) -- Kevin Brubeck Unhammer GPG: 0x766AC60C
signature.asc
Description: PGP signature
------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/
_______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
