Modelsofnatural language strongunderstandingstrong - PNAS.pdfVIP

Proc. Natl. Acad. Sci. USA Vol. 92, pp. 9977-9982, October 1995 Colloquium Paper This paper was presented at a colloquium entitled Human-Machine Communication by Voice, organized by Lawrence R. Rabiner, held by the National Academy of Sciences at The Arnold and Mabel Beckman Center, in Irvine, CA, February 8-9, 1993. Models of natural language understanding MADELEINE BATES BBN Systems and Technologies, 70 Fawcett Street, Cambridge, MA 02138 ABSTRACT This paper surveys some of the fundamental Makhoul and Schwartz (1), Jelinek (2), Levinson (3), Ober- problems in natural language (NL) understanding (syntax, teuffer (4), Weinstein (5), and Wilpon (6) attest to this fact. semantics, pragmatics, and discourse) and the current ap- But it is important to distinguish language understanding proaches to solving them. Some recent developments in NL from recognizing speech, so it is natural to ask, why the same processing include increased emphasis on corpus-based path has not been followed in natural language understanding. rather than example- or intuition-based work, attempts to In natural language processing (NLP),as we shall see, there is measure the coverage and effectiveness ofNL systems, dealing no easy way to define the problem being solved (whichresults with discourse and dialogue phenomena, and attempts to use in difficulty evaluating the performance of NL systems), and both analytic and stochastic knowledge. Critical areas for the there is currently no general way for NL systems to automat- future include grammars that are appropriate to processing ically learn the information they need to deal effectively with large amounts of real language; automatic (or at least semi- new words, new meanings, new grammatical structures, and