Proceedings of the 5th conference on Computational linguistics -

Abstract

The present collection of papers covers a wide range of interests and attitudes. While presumably nobody will enjoy every item, I expect every reader to fred several contributions which are stimulating to him and every contribution to find delighted readers. Since I am one of the few who have read them all - acting as the Chairman of the Program Committee - let me say a few words on the criteria adopted in organizing this work. The characteristic feature of Computational Linguistics is a focus on computation, on the derivation of results by a &laquo; mechanical &raquo; procedure, operating according to rules, according to an &laquo; algorithm &raquo;. A good tool for computation is, in malay cases, a computer, but computational linguistics is not the same as Computer-based Linguistics or Linguistic Data Processing (<i>Linguistische Datenverarbeitung</i>). Firstly, much computation can be and has been performed without computers. The choice of tools in, say, a statistical study is naturally a question of economy, not of scientific method. Secondly, several studies of methodology, aiming at formulating the linguistic problems in a manner which makes them amenable to computation, do not involve actual computation, let alone the direct use of computers. The demarcation against other mathematical approaches to linguistics is necessarily unclear: for some papers on design of models, the emphasis on problems of computability is not indisputable. Thirdly, linguistic research, like investigations in so many other fields, is often aided by the services of a computer without being, on that account, directed towards problems of computation. Thus lexicographic work is neither more nor less computational because the clerical part of it has become easier - or possibly more complicated - thanks to new equipment. The data processing performed in linguistic institutes of various kinds is certainly woi:th studying in its own right - preferably together with experts of economy, organization and office rationalization - but does not constitute a separate branch of scientific research. Again, the distinction is ofteri vague in practice, and the Program Committee is aware that some papers may have interest as reports on computational tools for linguistics rather than as contributions to Computational Linguistics proper. The Program Committee was careful and open in this respect, because of the intense interest in and need for such tools, at this point of time, in some fields. Fourthly, Computational Linguistics does not, in our view, include the study of linguistics-based information processing in Information Retrieval Systems, Programming Language Compilers and other cases where linguistics is applied as an aid to computation, unless such study is concentrated on linguistically crucial problems. The Program Committee was not in the position of initiating and controlling the research presented at the Conference. The authors obviously do not all share our view of Computational Linguistics as presented here. The Program Committee, by grouping the papers under a few headings, has tried to emphasize the various computational aspects, in a few cases more strongly than the authors have done themselves. It goes without saying, that many contributions could reasonably have been placed under more than one heading: semantical and syntactical analysis are often interlaced, and particularly so in some recent efforts; the meaning extraction procedures of Information Retrieval Systems come close to other semantic computations; Model Design is a component of very many contributions, and it is not intended as a ranking if that aspect has been considered more dominating in some papers. In short, the grouping under these headings is intended to characterize the field rather than the individual contributions. Although the Program Committee in its arrangements of the papers for the Conference and hence for this publication did not obediently adapt itself to the contents of the papers but tried to influence the discussions at the Conference and the influx of papers at the next COLING Conference in an intended way, this arrangement was decided <i>a posteriori</i>, on the basis of the material submitted. The choice of headings reflects the over-all trends in the field. Conspicuously, there is a large number of papers on what we have called Semantic Computation. Not many years ago, that title might have sounded as <i>contradictio in adjecto</i>. The inclusion of Semantics into the domain of the computable seems to mark a major progress in recent linguistics.