British Computer Society Natural Language Translation Specialist Group Web-site: http://www.bcs.org.uk/siggroup/sg37.htm

Machine Translation Review
Issue No. 11, December 2000 - pages 27-28
ISSN 1358-8346

This page URL: http://www.bcs.org.uk/siggroup/nalatran/mtreview/mtr-11/mtr-11-11.htm
Size: 3 A4 pages when printed

 
Book Review
Reviewer: Nadia Mckendrick

Breadth and Depth of Semantic Lexicons
Evelyne Viegas (ed.) (1999)
Dordrecht/Boston/London: Kluwer Academic Publishers. Hardback,
ISBN 0-7923-6039-7

The volume under review covers the main topics of the workshop on Breadth and Depth of Semantic lexicons, held at the university of California, in conjunction with the meeting on the 34th Annual Meeting of the Association for Computational Linguistics, on June 28, 1996, Santa Cruz, California.

Most of the books about computational lexical semantic lexicons deal with the depth (or content) aspect of lexicons, ignoring the breadth (or coverage) aspect. This book presents a first attempt to address both issues: depth (or content) and breadth (or coverage) of computational semantic lexicons, including lexical and conceptual approaches. Moreover, it addresses issues that have not yet been tackled in implemented systems such as the application time of lexical rules. This book embraces several different fields: Linguistics (theoretical, computational), semantics, pragmatics, psycholinguistics, cognitive science, computer science, artificial intelligence, knowledge representation, statistics, and natural language processing.

This book addresses two main issues of capturing regularities in the lexicon:
1. the treatment of lexical ambiguity;
2. lexical rules as a conceptual tool for controlled proliferation of entries.

Whereas the former has been regarded as a topical issue for quite some time, the latter is only now receiving its due attention. This book addresses both issues with a focus on lexical rules as a regulator of breadth and depth of the lexicons.

The book also investigates whether theoretical accounts of the use of lexical rules are too general and underspecified to support actual processing or not. Practical issues connected with the concrete application of lexical rules are also investigated: when to apply the rules; how the rules influence system design; and how to re-examine and adjust the theoretically posited rules in view of practical needs and evidence.

The book under review comprises four sections: lexical rules and underspecification; breadth of semantic lexicons; depth of semantic lexicons; lexical semantics and pragmatics. All these issues are discussed in fifteen chapters via computational and descriptive case studies addressed within the framework of a theory and an implemented system.

 
Section 1: lexical rules and underspecification

The chapters of this section discuss the following:
1. The different types of lexical rules an NLP (natural language processing) system could use (B. A. Onyshkevich). There are three sets of parameters a system developer needs to take into consideration when starting building a practical NLP system: the first one addresses the scope of linguistic phenomena covered by the lexical rule (LR) mechanism, including derivational word formation, inflectional morphology, and regular polysemy. The second parameter addresses the application of LRs in a computational system (at lexical acquisition time, lexicon load time, or at run time). The third parameter addresses the triggering and constraining mechanisms of LR applicability needed to avoid over-generation. The categorisation is also used to overview and classify the approaches to LR in the other papers of this book.
2. Analysis of present studies on the advantages and disadvantages of using lexical rules to extend a lexicon (T. Sehitoglu, C. Bozsahin), outlining a lexical organisation for Turkish that makes use of lexical rules for inflections, derivations and lexical category changes to control the proliferation of lexical entries. Lexical rules handle changes in grammatical roles; enforce type constraints and control the mapping of subcategorisation frames in valency-changing operations. Formally, constraints are enforced via a lexical inheritance hierarchy. The design has been tested as part of an HPSG grammar for Turkish. Concerning the application time of the rules, the authors, based on their experience, reckon that a run-time execution of the rules seems to be a far better alternative than precompilation in view of the intensive use of inflections and derivations in Turkish.
3. B. Gillon's chapter investigates the systematic connection between English mass and count nouns. The author addresses the principal morphological and semantic properties of the mass count distinction; the empirical generalizations pertinent to systematic connection between English mass and count nouns, in terms of lexical rules; and how such rules fit with a syntactic and semantic theory of common English noun phrases.
4. A. Sanfillippo's chapter presents an alternative approach to LRs to treat lexical ambiguity. The author suggests that regularities about sense/usage extensibility should be represented via underspecification in the lexicon. Lexical ambiguity is expressed by associating each ambiguous word form with an Underspecified lexical type which subsumes a class of alternations describing all admissible uses of the word. Appropriate lexical usage can then be assessed deterministically by utilizing syntactic and semantic cues gathered during text processing from the local phrasal context in order to ground underspecified word entries.

 
Section 2: breadth of semantic lexicons

This section deals with the large-scale acquisition of computational semantic lexicons using lexical rules or concept rules as a conceptual tool to extend a lexicon and existing knowledge sources, such as Word Net, Levin's database of alternation verbs and machine readable dictionaries. Some authors gave preliminary results on the evaluation of their approach with respect to word sense disambiguation (B. J. Dorr and D. Jones). The chapter on deverbal adjectives illustrates how lexical rules can be used to proliferate entries. The authors argue for a large scale approach when using lexical rules (V. Ruskin and S. Nirenburg). The usefulness of lexical rules as in practical system is being questioned, as it seems there are more irregularities than can be managed by their mechanism. (K. J. Burns and A. R. Davis).

 
Section 3: depth of semantic lexicons

The first two chapters focus on semantic phenomena among adjectives (P. Bouillon), nouns and nominal compounds (M. Johnston and F. Busa). The authors here are more interested in investigating the depth of the phenomena they cover rather than the breadth aspect, although scalability is also addressed by some authors. The last two chapters deal with automating the acquisition of semantic lexicons (J. N. Chen and J. S. Chang) and a corpus of free- responses and lexical semantic techniques (J. Burstein, S. Wolf, C. Lu). These authors give preliminary results on the evaluation of their approach with respect to word sense disambiguation for the former, and information extraction for the latter.

 
Section 4: lexical semantics and pragmatics

The chapters in this section deal with various perspectives on the role of lexical semantics and pragmatics. The first chapter illustrates the Word Net approach to lexicon organization comparing its approach to a more syntactic-based approach as described in Levin (1993), showing the absence of an isomorphism between syntax and semantics (C. Fellbaum). The second chapter investigates a new way to get around the incompleteness of computational lexicons by using both lexical rules and conceptual rules to create new lexicon entries and new concepts on the fly (E. Viegas). The third chapter discusses two main approaches to (computational) lexical semantics, Supply-side and demand-side, and primary focusing on theoretical and practical aspects in computational linguistics (S. Nirenburg and V. Ruskin). The last chapter discusses lexical rules and lexical underspecification. The authors argue for pragmatic rules of conversation along with contextual knowledge to treat lexical polysemy (S. Helmreich and D. Farwell).

 
References

Leech, G. (1981) Semantics, Cambridge University Press.
Levin, B.(1993) English Verb Classes and Alternations: A Preliminary Investigation, University of Chicago Press