[lkb] LKB morphology

Jesse Tseng Jesse.Tseng at loria.fr
Wed Apr 18 19:55:38 CEST 2007


I am attempting to implement a version of Bonami and Boyé's
(2002, from HPSG Trondheim) treatment of French verb stem
alternation using the LKB morphology tools.

This involves listing several distinct stems as part of each
verb's lexical entry, and then letting the suffixation rules
pick the stem they need to add their ending to.

For example, the verb "boire" lists at least 4 different stems: 
   boi (bois, boit, ...)
   boiv (boivent, boives, ...)
   bu (bus, bûmes, ...)
   buv (buvons, buvant, ...)

In the lexicon, the ORTH value of the verb would just be left
uninstantiated, or else filled in with a citation/default/dummy
form and then modified (not by an affixation rule, but by a
lexical rule that simply replaces the entire value of ORTH, e.g.
"boire" --> "buv", "pouvoir" --> "puiss").

I don't think the LKB's orthographemic processor will easily
allow either of these approaches, because ORTH really needs to
be specified lexically, and the surface forms have to be related
to this underlying ORTH by prefixation/suffixation only. Is this
correct? 

(Going through "irregs.tab" is not a reasonable solution for
French, especially since the goal is to stay as faithful as
possible to BoBo's analysis).

Has anyone else tried something like this?

Thanks for any suggestions!

Jesse



More information about the lkb mailing list