[developers] Trigger rules with specialization
danf at stanford.edu
Tue Dec 13 20:08:59 CET 2011
Hi Woodley -
Yes, the use of two lexemes for "do" and "does" is independent of generation, and has two motivations, just FYI: (1) even though the ERG does not currently contain phonological information about each lexical entry (beyond the vocalic/consonantal onset property), it should someday, and the pronunciation of "does" is idiosyncratic, so if we didn't use a separate lexeme, we would somehow have to specify this surprising phonology in an extension of the irregs file mechanism, associated with the v_3s-fin_olr inflectional rule; and (2) the inflectional characteristics of auxiliary verbs are varied enough that it has seemed easier to simply define the fully inflected forms for them directly in the lexicon, rather than try to coerce them into the regular verb inflectional rules.
As for your proposed extension of the trigger rules, I would welcome the additional expressive power, and would be glad to invest some time in revising the current inventory of trigger rules, confident that we could see some significant efficiency gains in generation. My only caveat is that it would be good to at least amend the LKB's reader for the trigger rules so that it wouldn't trip over the additions, even if it isn't ready to use them, so the grammar writers can still maintain just one version of the rules for both generators.
----- Original Message -----
From: "Woodley Packard" <sweaglesw at sweaglesw.org>
To: developers at delph-in.net
Sent: Saturday, December 10, 2011 10:44:55 AM
Subject: Re: [developers] Trigger rules with specialization
Regarding the English example I gave -- whoops! excuse me, I had meant to say: one for "do" and one for "does". And I was wrong -- the current ERG does not actually avoid allowing both forms into the generation chart, although it perhaps could(?); the trigger rules for these two forms are nearly identical. There was presumably some other good reason to separate those lexemes in this case. In any case, the point stands for the other languages I mentioned.
On Dec 10, 2011, at 10:04 AM, Woodley Packard wrote:
in English, the auxiliary verb forms "did" and "does" are vacuous, but their inflection depends on the agreement properties of their subject. The ERG avoids allowing both forms into the generation chart by having two separate lexemes and two separate trigger rules -- one for "did" and one for "does"
More information about the developers