[Prev][Next][Index]

XTAG meeting on May 11, 10:30am


Hi all, 
since we cancelled the xtag meeting last week due to the
unexpected power cut, Anoop will give his practice talk this Thursday. In
addition, Fei will also talk on the comparison of XTAG with the Penn
Treebank grammar.

Both abstracts are attached below:



----------------------------------------------------
Practical Expts in Parsing using TAGs - Anoop Sarkar
----------------------------------------------------

We report on some practical experiments in parsing using lexicalized
Tree Adjoining Grammars. We parse 2250 sentences from the Wall Street
Journal using this parser. In these experiments we measure observed
parsing times taken by the parser to produce all valid parses for each
sentence in the form of a shared derivation forest. The results
suggest that the observed complexity of parsing for LTAG is dominated by
factors other than sentence length.


-----------------------------------------------------------------
Compare the XTAG with the English Penn Treebank grammar - Fei Xia
-----------------------------------------------------------------

We will compare the XTAG grammar with a Treebank grammar (Ext-G) extracted
from the English Treebank. About half of the templates in the XTAG grammar
"match" some templates in Ext-G, which account for about 82% of template
tokens in the Treebank. The remaining 18% of template tokens are covered
by 2700+ templates in Ext-G. We manually examined 306 of these templates
to decide why they are missing in the XTAG grammar. We conclude that the
majority of mismatches in the two grammars are due to different analyses
to the same phenomena. Finally, we will also go through some constructions
that seem to be missing from the XTAG grammar.



--------

- Rashmi