[developers] How do I make PET deal with unknown lexical items?
Bill McNeill (UW)
billmcn at u.washington.edu
Wed Jul 30 01:30:53 CEST 2008
I'm just starting to use PET. I've got it parsing simple sentences using
the July 2007 ERG. Now I'm trying to get it to deal with sentences that
contain lexical items it doesn't recognize. I know there are ways to
specify richer input information like POS tags on the words, but I haven't
been able to get this to work. I'm looking for pointers to more
documentation or better yet input files I could work with.
Say I try to parse a sentence like "Xavier is tall." PET/ERG 2007 can't
handle this because there is no lexicon entry for Xavier. I want to be able
to tell PET that "Xavier" is a proper noun and have it generate parses. The
first thing I've tried is using the YY-input mode as documented on the
PetInput <http://wiki.delph-in.net/moin/PetInput> page. I copied the
example sentence "This FAQ will concentrate..." with its YY annotation from
that page, put all the elements on a single line, and ran it in PET using
the command line
cheap -yy -packing -verbose=4 -default-les english.grm
as specified on that page. However, this gave me the following error
no lexicon entries for:
"faq" [NP1 ]
"." [. ]
1. Can anyone send me a simple YY-input format file that is known to work
with the July 2007 ERG and a cheap command line that will parse it?
2. Is there documentation for the YY and XML-based PIC formats anywhere
other than the PetInput page on the Delphin Wiki? I'm still unclear where
I'm supposed to be looking for documentation.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the developers