This workshop will evaluate the need for computational dictionaries, the value of large text databases in creating them, and the role of cooperative efforts in this area. Both commercial and scientific interests will be considered. By "computational dictionaries" We encompass databases containing information about the form, pronunciation, meaning and usage of words and their collocations. Such databases are not merely processed versions of conventional dictionaries. Examples range from a relation among company names, and product types; to a relation among verbs, derived nominals, and their co-occurence probabilities with various argument structures. We believe that the quality of computer programs for manipulating natural language depends crucially on the quality of such word-indexed databases, and that theories of the representation and manipulation of lexical information are central to scientific theories of language structure and language use. We also believe that some crucial kinds of lexical information can best be discovered by computer processing of very large bodies of text. Progress in research and in its commercial exploitation is now hindered by the unavailability of suitable computational dictionaries, and of generally-available text databases that are large enough to help construct them. This workshop will try to arrive at a consensus about what lexical and textual resources are most needed, what mix of co-operative and individual efforts is appropriate to generate them, and what organizational structure is appropriate to coordinate efforts and distribute results. It will produce a report reflecting the consensus of the participants.
