Bunt H., Muskens R. (eds.) Computing Meaning. Volume 1

Издательство Kluwer, 1999, -371 pp.Computational semantics is concerned with computing the meanings of linguistic objects such as sentences, text fragments, and dialogue contributions. As such it is the interdisciplinary child of semantics, the study of meaning and its linguistic encoding, and computational linguistics, the discipline that is concerned with computations on linguistic objects.
From one parent computational semantics inherits concepts and techniques that have been developed under the banner of formal (or model-theoretic) semantics. This blend of logic and linguistics applies the methods of logic to the description of meaning. From the other parent the young discipline inherits methods and techniques for parsing sentences, for effective and efficient representation of syntactic structure and logical form, and for reasoning with semantic information. As is often the case with inheritance, blessings are mixed and computational semantics' legacy not only contains useful elements but also many problems and unresolved issues. In fact, many of the well-known problems in semantics such as those relating to handling ambiguity and vagueness, as well as fundamental issues regarding compositionality and context- dependence, re-appear in computational semantics in a form even more compelling than in formal semantics. In this introductory chapter we shall highlight some of the issues and problems we think are most pressing. We shall start with discussing a straightforward way to implement formal semantics in the tradition that started with Montague's work. Then, after a conceptual analysis of the principle of compositionality and a discussion of its role in computational work, the problem of ambiguity will be introduced and it will be shown that the straightforward way of taking standard formal semantics as a specification of computer algorithms needs to be amended. An important issue in computational semantics is that of the context dependency of language. The task set to computational semanticists is to provide concrete linguistic objects with abstract meanings; as concrete linguistic objects are utterances rather than sentences (types of utterances), computational semantics needs to draw in a lot of the work that takes context into account and traditionally goes under the banner of pragmatics. After a discussion of this, and of the need for underspecified representations of meaning, which arises out of the ambiguity problem, the chapter closes with an overview of the other chapters in this book.Computational Semantics
On Semantic Underspecification
Dynamic and Underspecified Interpretation without Dynamic or Underspecified Logic
Labeled Representations, Underspecification and Disambiguation
Underspecified Semantics in HPSG
Minimum Description Length and Compositionality
How to Glue a Donkey to an f-Structure: Porting a 'Dynamic' Meaning Representation Language into LFG's Linear Logic Glue-Language Semantics
Vague Utterances and Context Change
Using Situations to Reason about the Interpretation of Speech Events
Simulative Inference in a Computational Model of Belief
Indefinites as Epsilon Terms: A Labelled Deduction Account
Dynamic Skolemization
Semantically-based Ellipsis Resolution with Syntactic Presuppositions
Presupposition Projection as Proof Construction
Dynamic Discourse Referents for Tense and Modals
Linking Theory and Lexical Ambiguity: The Case of Italian Motion Verbs
A Disambiguation Approach for German Compounds with Deverbal Head