Thank you! I'm beginning to understand. --snip--
> > > > Francis writes: > > "The lexical selection is done in the aptly-named "lexical selection" > > stage, which sits between lexical transfer (which outputs all the > > possible translations of each word) and structural transfer" > > > > But most alternative translations are already discarded by the tagger, > > aren't they? In the very first step. > > No, that's alternative morphological analyses. > > Plain lt-proc (morphological analysis) gives one or more analyses: > > ^bank/bank<n><m><sg><ind>/bank<vblex><imp>$ > > apertium-tagger chooses one _analysis_: > > ^bank/bank<vblex><imp>$ > > lt-proc -b (lexical transfer) adds one or more translations (and retains > the original one): > > ^bank<vblex><imp>/beat<vblex><imp>/knock<vblex><imp>$ > > lrx-proc (lexical selection) chooses one _translation_ from these: > > ^bank<vblex><imp>/knock<vblex><imp>$ > > and then apertium-transfer moves words around or whatever. > > > Of course, if bank<n> were chosen by the tagger, that would also lead to > a different translation, but it's a choice of a different sort (is it a > noun or is it a verb, rather than what translation does that particular > verb have). > > -- > Kevin Brubeck Unhammer > > GPG: 0x766AC60C > OK. What I didn't understand, was what the lexical selection module had to choose between. As far as I could see, there was only one alternative and not much info left for it to work with. I forgot that the lexical transfer actually adds info. Per ------------------------------------------------------------------------------ Free Next-Gen Firewall Hardware Offer Buy your Sophos next-gen firewall before the end March 2013 and get the hardware for free! Learn more. http://p.sf.net/sfu/sophos-d2d-feb _______________________________________________ Apertium-stuff mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/apertium-stuff
