City University of Hong Kong Dep
Department of Chinese, Translation and Linguistics
The Halliday Centre for Intelligent Applications of Language Studies
From Montague to Semantic Annotation
Prof. Harry Bunt
Professor of Language and Information Science,
Tilburg University, the Netherlands
Date: 12 Jun 2008, Thursday
Time: 4:30pm - 6:00pm
Venue: B7603, (Lift 3, 7/F, Blue Zone), Academic Building, CityU
The question how the meaning of natural language expressions can be computed has over time fascinated philosophers, logicians, linguists, psychologists, and computer scientists (in particular researchers in Artificial Intelligence). The philosopher Ludwig Wittgenstein proposed a view on meaning that was formalized by the logician Alfred Tarski and taken further by Richard Montague (and later also by Hans Kamp), and that formed the basis for the establishment of formal semantics as a branch of linguistics, bordering on mathematical logic. Two assumptions were fundamental for this development to be possible: the interpretation of sentence meaning as a proposition's truth conditions, and the principle of compositionality. The formal specification of algorithms for generating semantic representations of natural language sentences in intensional logic by Montague in the seventies, and those for building first-order logic representations for natural language texts by Kamp in the eighties, form the highlights of this approach to the computation of meanings.
Computer programs implementing the Montague and Kamp semantics approaches have not been very succesful, mostly for the following three related reasons: the ubiquity of ambiguity, the context-dependence of language use, and the importance of nonlinguistic modalities in everyday use of language. Computational studies have revealed the massive character of ambiguity in natural language, which is even greater than many linguists are aware of. Researchers often regard this as a problem, and a bad property of natural language. I will argue instead that ambiguity is one of the most important sources of the power of natural language as an instrument of communication, and that ambiguity exists only when context and nonlinguistic modalities are neglected. On the other hand, the context-dependence of meaning in natural language challenges the principle of compositionality, which is at the basis of all algorithms for computing meaning in a systematic fashion.
The essence of all attempts to deal with this situation is to combine information from the context in which the natural language expression occurs which the available linguistic semantic information to compute its meaning. In this talk I will discuss some of these attempts, in particular those concerned with the use of underspecified semantic representations and those which start from newly developing approaches to semantic annotation.
Harry Bunt is professor of Language and Information Science at Tilburg University, the Netherlands, and Head of the Department of Communication and Information Sciences in the Faculty of Humanities. He was programme director of the Dutch national research programme 'Human-Computer Communication Using Natural Language', director of the inter-university research programme 'Dialogue Management and Knowledge Acquisition', and adviser of the German Ministry of Research and Technology for language technology. He is chairman of the ACL Special Interest Group in Parsing, secretary of the ACL Special Interest Group in Computational Semantics (SIGSEM), chairman of the SIGSEM Working Group on the Representation of Multimodal Semantic Information, and member of the Board of the Dutch National Graduate School in Logic, and secretary of the Dutch Organisation for Language and Speech Technology. He is also chairman of the ISO expert group on semantic content, co-project leader of the project on the standardization of temporal information annotation, and project leader of the project on developing standards for dialogue annotation. He has published and edited a number of books, including 'Computing Meaning, Volumes 1-3' (1999, 2002 and 2007), 'New Developments in Parsing Technology' (2004), 'Cooperative Multimodal Communication' (2001), 'Abduction, Belief and Context in Dialogue Studies in Computational Pragmatics' (2000), 'Advances in Probabilistic and Other Parsing Technologies' (2000), 'Multimodal Human-Computer Interaction' (1998) and 'Mass Terms and Model-Theoretic Semantics' (1985).
~ All Are Welcome ~
CTL Technical Support