[itsdb] i-category and phenomena classification

Emily M. Bender ebender at u.washington.edu
Sun Apr 9 01:20:42 CEST 2006


Hi folks,

I've got students creating test suites to use with the fine system in
my grammar engineering class this quarter, and one of the pieces of
information we're recording is the phenomena illustrated by each
example.  Of course, some examples illustrate multiple phenomena.

It seems that this information is expect to be stored in the
i-category field (though that's not entirely apparent on the
ItsdbReference page on the wiki).

My question is whether there is any support for multiply-classified
examples.  If I create complex labels for that field, is it possible
to display/aggregate over examples whose i-category contains a certain
substring?

If not, what is the canonical way of handling examples which
cover multiple phenomena?

[NB: Laurie Poulson has pointed out that most references on test suite
construction specifically say to avoid examples which illustrate
more than one phenomenon.  One the one hand, this isn't strictly
speaking possible: every sentence has some word order, some case
(in languages with case) etc.  On the other hand, we don't think it's 
advisable: While one surely wants the simplex sentences illustrating
one thing at a time, the multiple-phenomena sentences are crucial
for testing interactions between analyses.]

Emily



More information about the itsdb mailing list