[Prev][Next][Index]

XTAG meeting on Thursday, 21 Sept., 5pm (NOTE TIME CHANGE!)


Hi all,

First of all, please note that we will be meeting at a DIFFERENT TIME this
Thursday. There is a talk at 10:30 in the anthropology colloquium
series that many of us would like to attend, so we'll be meeting instead
at:

5:00pm in the LARGE CONFERENCE ROOM.

Yusuke Miyao and Naoki Yoshinaga will be giving two talks. They are IRCS
short term visitors from Tokyo University. Their abstracts follow:

------------------------------------------------------------------------
(1) HPSG-based systems - Yusuke Miyao
--------------------------------------

Abstract:

This talk will give you an overview of the HPSG-based systems developed in
Tsuj ii group, Tokyo University.  Our group aims at developing basic
technologies for intelligent natural language processing, that is,
linguistically sound as well as practical enough for real-world
applications.  The research succeeded in making various HPSG-based
systems, which are efficient programming environment, HPSG parsers,
wide-coverage HPSG grammars, statistical disambiguation methods, and their
application to information extraction.  I will describe advanced
technologies in each of them, and how they achieve efficiency,
wide-coverage, and a ccuracy.  I also demonstrate our English grammar,
XHPSG, translated from the XTAG English Grammar.


---------------------------------------------------------------------
(2) Automatic conversion from FB-LTAG into HPSG - Naoki Yoshinaga
------------------------------------------------------------------

Abstract:

I talk about an automatic conversion algorithm of Feature-Based
Lexicalized Tree Adjoining Grammar (FB-LTAG) into Head-driven Phrase
Structure Grammar (HPSG). LTAG elementary trees are converted into lexical
entries of the target grammar, by regarding the leaf nodes (both
subsitution nodes and foot nodes) as the arguments of the anchors and
encoding them into a stack. The two tree operations, substitution and
adjunction are emulated by grammar rules, which manipulates the stack. We
applied this algorithm to the XTAG English grammar and preliminary
experimental results show that the obtained grammar generates trees
strongly equivalent to the ones generated by the original grammar.

***