[logon] Mysterious problem with regexps in transfer rules
Stephan Oepen
oe at ifi.uio.no
Sat Nov 15 15:30:27 CET 2008
hi berthold (copying the `logon' list, in case others are interested),
> I discovered a mysterious problem with regexps today.
>
> When testing my transfer grammar interactively (form Browse results)
> the following transfer rule happy applies and I can generate from the
> resulting MRS.
>
> jeder_every_q_mtr := quantifier_mtr &
> [ INPUT.RELS < [ PRED "~(de:_jed_q_rel|de:_fast[+]jed_q_rel)" ] > ,
> OUTPUT.RELS < [ PRED _every_q_rel ] > ].
>
> However, if I batch test in [incr tsdb()] (using the transfer grammar
> in the same lisp process), I get an unknown transfer predicates error
> message:
[...]
> Of course, I could split the rules, but a concise representation would
> be nice.
i am impressed you do consider "~(de:_jed_q_rel|de:_fast[+]jed_q_rel)"
concise :-). but i am afraid the behaviour you report is `correct': at
the start of transfer, for rule sets that were loaded with `:filter t',
all input predicates are compared (as strings) against the `dictionary'
of known INPUT predicates in that rule set. that test, currently, does
not take into account regular expressions (though it probably should).
a way of working around this limitation could be to normalize predicate
names in the first (GG accomoation) phase, and turn off filtering there
(relatively few rules, not much of a loss). this is what we have been
doing for the NoEn LOGON system, so i never noticed the limitation :-).
but for this to work, `_jed_q' and `_fast+jed_q' have to be considered
equivalent, at least for transfer purposes (if you can translate both
to `every', you might be willing to make that assumption). otherwise,
i see no alternative currently to breaking the rule into two.
all best - 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 logon
mailing list