[developers] SOLVED: Re: Mysterious generation problem in LOGON
Ann.Copestake at cl.cam.ac.uk
Sun Feb 10 01:03:32 CET 2008
yes, sorry, didn't read carefully enough. OK, well, there are
possible situations where one might perhaps want to do this in some
sort of construction semantics, but we're essentially excluding such
cases with this generator algorithm, so indeed, a good idea to warn.
> > indexing on predsort would not be slow with a small grammar and it's
> > exactly the kind of thing I would like to be able to demo in that sort
> > of context. Perhaps if something warned when a sort was compatible
> > with `too many' lexical entries - there's no reason to single out
> > predsort.
> are you maybe thinking of the inverse situation, using underspecified
> predicates in generator inputs? there i can see how `predsort' would
> make sense, in principle. but, indeed, a warning about a very large
> number of lexical entries or rules activated in chart initialization
> would also seem potentially useful.
> but berthold had a rule in his grammar with C-CONT [ PRED predsort], or
> more likely no mention of PRED in the rule at all, such that `predsort'
> was there merely by appropriateness. i suspect such underspecification
> in rules or lexical entries will typically be unintended. as i fix the
> incorrect treatment of such cases in the current LOGON code, i can see
> whether there is an easy way to generate such warnings (in fact, i had
> thought we did warn about these cases).
> cheers - oe
> +++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
> +++ CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
> +++ --- oe at ifi.uio.no; oe at csli.stanford.edu; stephan at oepen.net ---
More information about the developers