Natural Language Research 
Aravind Joshi, Mitch Marcus, 
Mark Steedman, and Bonnie Webber 
University of Pennsylvania 
OBJECTIVE 
The m~n objective is basic research and system development leading to (1) characterization 
of information carried by (a) syntax, semantics, and discourse structure, (b) their relation to 
information carried by intonation, and (c) development of methods for using this information 
for generation and understanding; (2) development of architectures for integration of utterance 
planning with lexical, syntactic and intonational choice; (3) development of incremental strate- 
gies for using syntactic, semantic, and pragmatic knowledge in understanding and generating 
language. 
RECENT ACCOMPLISHMENTS 
• Completed an implementation of a parser for the Tree Adjoining Grammar (TAG). 
• Experimental evaluation of the TAG parser showed that the two-pass strategy in the 
current parser led to 90% reduction in the number of states and 85% reduction in the 
number of elementary trees, as compared to the one-pass strategy. 
• Developed a formal theory unifying intonational structure, discourse information struc- 
tures and surface syntactic structures within the categorial framework. 
• Extended earlier characterization of natural language tense and aspect in terms of event 
ontologies and discourse structure. 
• Learning- See adjoining report on "Large Corpus Development." 
• Completed a pilot implementation of the multi-functional cooperative response system. 
• Developed a novel approach for integrating syntax and semantics in TAGs, and explored 
its application to generation and translation. 
PLANS 
• Develop applications of categorial theory to the problem of computer synthesis of contex- 
tually appropriate intonation in spoken language. 
• Begin implementation of the theory of tense and aspect using an event calculus. 
• Complete the work on the lexicon for the TAG parser and test the parser. 
• Complete the development of a deterministic parser for TAGs. 
• Develop the formalism of synchronous TAGs (for integrating syntax and semantics), de- 
velop a parser for synchronous TAGs and explore further its application to machine trans- 
lation. 
• Learning -- See adjoining report on "Large Corpus Development." 
427 
