The Logic of Typed Feature Structures：（输入的逻辑功能结构）.pdfVIP

Computational Linguistics Volume 19, Number 3 The Logic of Typed Feature Structures Bob Carpenter (Carnegie Mellon University) Cambridge, England: Cambridge University Press (Cambridge Tracts in Computer Science 32, edited by C. J. van Rijsbergen), 1992, viii + 270 pp. Hardbound, ISBN 0-521-41932-8, $34.95 Reviewed by Fernando Pereira AT T Bell Laboratories 1. Introduction For those of us who belonged to the Bay Area (Computational) Linguistics Commu- nity, the early eighties were a heady time. Local researchers working on linguistics, computational linguistics, and logic programming were investigating notions of cat- egory, type, feature, term, and partial specification that appeared to converge to a powerful new approach for describing (linguistic) objects and their relationships by monotonic accumulation of constraints between their features. The seed notions had almost independently arisen in generalized phrase structure grammar (GPSG) (Gazdar et al. 1985), lexical-functional grammar (LFG) (Bresnan and Kaplan 1982), functional- unification grammar (FUG) (Kay 1985), logic programming (Colmerauer 1978, Pereira and Warren 1980), and terminological reasoning systems (Ait-Kaci 1984). It took, how- ever, a lot of experimental and theoretical work to identify precisely what the core no- tions were, how particular systems related to the core notions, and what were the most illuminating mathematical accounts of that core. The development of the unification- based formalism PATR-II (Shieber 1984) was an early step toward the definition of the core, but its mathematical analysis, and the clarification of the connections between the various systems, are only now coming to a reasonable closure. The Logic of Typed Feature St