LSTL 10: Lexical-semantic information in Head-driven Phrase


LSTL 10: Lexical-semantic information in Head-driven Phrase

Artikel-Nr.: ISBN 9783895865831
111,60
Preis inkl. MwSt., zzgl. Versand


 Lexical-semantic information in Head-driven Phrase

Structure Grammar and Natural Language Processing
Martin Hoelter
University of Bochum

Lexical-semantic information traditionally has certainly not been in the focus of formal grammar theories or computational linguistics. With the recent emergence of complex Natural Language Processing (NLP) systems, however, an immediate practical need for semantically richer but formalized language-related information has arisen. The question then is how e.g. selectional restriction information which normally has been dealt with by separate models and theories can be integrated in a formal grammar framework also adequate for NLP tasks.

The suitable grammar theory employed is Head-driven Phrase Structure Grammar (HPSG), whose complex feature structure models and attribute-value matrices as major means of representation have made it the linguistic framework of choice in state-of-the-art NLP. Cobuild dictionaries on the other hand provide the ideal lexicographic framework for a data base: their unique definition style using simple English sentences only is shown to be both of high theoretical and practical relevance.

"Lexical-semantic information in Head-driven Phrase Structure Grammar and Natural Language Processing" proposes a concept of information transfer between natural language dictionaries, formal grammar, and language engineering systems by mapping dictionary definitions to HPSG lexical entries. The aim is to show how retrieval of lexical-semantic information from a Cobuild dictionary can be organized and what the theoretical assumptions on lexical-semantic information in a syntactically and semantically integrative HPSG model are.

ISBN 9783895865831. LINCOM Studies in Theoretical Linguistics 10. 200pp. 1999.

Diese Kategorie durchsuchen: LINCOM Studies in Theoretical Linguistics (LSTL)