Difference between revisions of "Prefixes and infixes"

From Apertium
Jump to navigation Jump to search
(New page: <pre> <mlforcada> I haven't thought much about it but I think the solution would be <mlforcada> to see lexical forms as sets and not as sequences <spectie> sergio said something about 'car...)
 
Line 1: Line 1:
  +
One solution would be to see lexical forms (LFs) as sets and not as sequences. e.g. pl.n.kitabu or pl.kitabu.n would be the same (swahili)
<pre>
 
  +
<mlforcada> I haven't thought much about it but I think the solution would be
 
  +
Two possible ways:
<mlforcada> to see lexical forms as sets and not as sequences
 
  +
<spectie> sergio said something about 'cartesian product'
 
 
# normalization of some kind, perhaps specified with some rules
<spectie> would this require changing the 'translation as a stream' thing ?
 
 
# treating LFs as sets
<mlforcada> so that pl.n.kitabu or pl.kitabu.n would be the same (swahili)
 
 
# detecting lexical forms in the sequence they are generated by the tagger
<mlforcada> in fact vitabu would naturally be pl.kitabu.n
 
<mlforcada> I think we should only change transfer
 
<mlforcada> something that either normalizes it or deals it in whatever order
 
<mlforcada> so that one could deal with pl.n.kitabu
 
<mlforcada> in fact
 
<mlforcada> in basque
 
<mlforcada> it would be very nice to have this for verbs:
 
<mlforcada> dakarzu: you bring it ; nakarzu : you bring me
 
<spectie> yep
 
<mlforcada> the morpheme for object/absolutive comes first
 
<mlforcada> so there are two possible ways
 
<mlforcada> (1) normalization of some kind, perhaps specified with some rules
 
<mlforcada> (2) treating LFs as sets
 
<mlforcada> and a middle way
 
<mlforcada> (1.5) detecting lexical forms in the sequence they are generated by the tagger
 
<mlforcada> however
 
<spectie> i think sets is probably cleaner ... although is there any information lost in the loss of ordering ?
 
<mlforcada> I think tagger rules should also be changed
 
<mlforcada> and we could have moyogo discuss lingala morphology which is also prefix, word classes, etc. as swahiki is
 
<mlforcada> I like it that verbs come marked with the classs of their object
 
<mlforcada> sw: nilikisoma: I read it where -ki- is it for a ki- class noun as kitabu
 
<spectie> is this inflectional or derivational ?
 
<mlforcada> ni-li-ki-soma : I-past-[ki object] soma
 
<mlforcada> I think it is inflection
 
</pre>
 

Revision as of 15:03, 26 May 2007

One solution would be to see lexical forms (LFs) as sets and not as sequences. e.g. pl.n.kitabu or pl.kitabu.n would be the same (swahili)

Two possible ways:

  1. normalization of some kind, perhaps specified with some rules
  2. treating LFs as sets
  3. detecting lexical forms in the sequence they are generated by the tagger