berthold.crysmann at gmail.com
Mon Sep 11 15:21:50 CEST 2017
in the context of a grammar for French, I have just tried setting
infl-pos to the first list member to correctly treat MWEs like
"pomme(s) de terre" or "kilomètre(s) par heure".
Here's the code snippet:
(defun find-infl-pos (unifs orths sense-id)
(declare (ignore unifs orths sense-id))
It seems that the above user-defined function works in parsing, but
does not work in either generation or when interactively applying
lexical rules to word/lexical entries.
Similarly, when setting the function to nil, parsing accepts with
inflection in either place, but generation still behaves according to
LKB's default, which is (length orths).
Grep'ing through the code, I find reference to infl-pos in
main/parse.lsp and main/lex.lsp, but not e.g. in mrs/generate.lsp,
which is sort of congruent.
What is the current status of the infl-pos business in the LKB. Having
divergent behaviour in parsing and generation does not seem optimal to
Furthermore, what do other processors do with inflection if *orth-path*
(or its equivalent) is a list?
Thanks in advance for clarifying this.
It appears to me that in the limiting case infl-pos should be
specifiable as a property of specific lexical entries.
Berthold Crysmann <berthold.crysmann at gmail.com>
CNRS, Laboratoire de linguistique formelle (UMR 7110)
Adresse postale : Case 7031, 5 rue Thomas Mann, 75205 Paris cedex 13
Adresse géographique : bureau 545, bât. Olympe de Gouges, 8 rue
Albert Einstein, 75013 Paris
More information about the developers