Linguistic Databases

Contents Introduction vii Bibliography 1 v Introduction This is a selection of papers on the use of databases in linguistics. All of the papers were originally presented at a conference entitled \Linguis-tic Databases", held at the University of Groningen March 23-4, 1995. This introduction reviews the motivation for a special examination of linguistic databases, introduces the papers themselves, and then, brieey, suggests both how the knowledge is useful to working linguists, as well as how databases might evolve to be used for linguistic data more easily. Motivation Linguistics is a data-rich study. First, there is a great deal of purely linguistic data. Linguistics sets itself the task of describing and analyzing the structure of language as this is evidenced in billions of speakers, each making hundreds of utterances a day for linguistic lifetimes of several decades. Factors important to the structure of these utterances include thousands of languages, each with tens to hundreds of thousands of words, which in turn may be found in dozens to hundreds of diierent word forms. The results of linguistic analysis show great variety in the particular coding rules at all levels|sounds, words and phrases. The rules are not individuated clearly enough to allow reliable comparisons, but they seem to number in the thousands in current formulations. The many factors studied in linguistic variation including geography, sex, social and educational status, pathology, and situational \register" (as for telegrams, graati, etc.) only add to this embarassment of riches. Second, various subbelds of linguistics have developed experimental methodologies involving a great deal of data which, although not purely linguistic, are used crucially in linguistic theorizing or in applications. Some examples of these would be physcial measurements such as air pressure and pitch, records of social and geographical parameters of variation, the quasi-linguistic ill-formed examples of generative vii viii / Linguistic Databases grammar, or psychological measurements such as reaction time or the movements of eyes in reading. Third, applications of linguistics need to keep track of further data categories such as successful (and unsuccessful) processings, degree of ambiguity in results, use of particular knowledge sources and heuristics, user reactions, and comparisons to alternative system conngurations. Given this amount of data, it is not surpising that a good number of linguists have investigated software for managing data. Databases have long been standard repositories in phonetics (see Liberman 1997, UCLA Phonetics Laboratory 1996) and psycholinguistics (see MacWhinney 1995) research, but …