Linguistics and Automated Language Processing

0.1 This paper is concerned with natural language, computers, and two groups of people interested in natural language: ]inguists, and persons engaged in computer processing of natural language data. There is some intersection of the latter sets, but the intersection is quite small relative to the size of the sets themselves and is thus inadequate to provide linguists with a proper perspective on automated language processing, or computer scientists with a proper perspective on linguistics. Although both groups of persons have a mutual interest in natural language, their conceptualizations of the nature of language and their approaches to processing language data are very different. To present a somewhat oversimplified view of these differences: linguists tend to be theory-oriented-they are concerned with interesting but sometimes quiteesoterlc problems, counterexamples , and the infinite set of sentences of competence; on the other hand, persons engaged in automated language processing tend to be data-oriented, and are concerned with statistical significance and with some finite subset of the sentences of performance. The question therefore arises as to whether these different perspectives are to be interpreted as incompatible or comple-r:~entary, and if complementary, whether some research concept might provide the means for a unified approach to analysis of natural language. In this paper, Section 1 deals with the perspective of linguists on automated language processing and computer scientists on linguistics; Section Z discusses their respective concepts of natural language and lI arn indebted to Paul GarTin for his valuable comments on this paper.