Machine-Readable Dictionaries and Computational Linguistics Research

Abstract

A substantial amount of work has gone into attempts to provide computational solutions to a wide range of problems which rely on information extracted from machine-readable dictionaries (MRDs). Syntactic parsing, grammar development, word sense selection, speech synthesis, robust text interpretation, knowledge acquisition, information management, phonetically guided lexical access, phrasal analysis — these are only a few of the language processing functions which have been substantially aided by the availability of large dictionaries on-line. The paper looks at representative examples from this list and analyses them within a framework which aims to establish broad categories within computational linguistics to which MRDs have been (or could be) applied, and modes of their use. In addition to questions to do with the nature of the coupling between the contents of a dictionary entry and the task to which this is applied, I will use the same case studies to address a class of issues at a different level of generality — these include the reliability of an MRD from an application’s point of view, the cost of extracting the information relevant to a computer program, and the optimal structure and organisation of a machine-readable source.