A Tool Kit for Lexicon Building

A TOOL KIT FOR LEXICON BUILDING Thomas E. Ahlswede Computer Science Department Illinois Institute of Technology Chicago, Illinois 6e616, USA ABSTRACT This paper describes a set of interactive routines that can be used to create, maintain, and update a computer lexicon. The routines are available to the user as a set of commands resembling a simple operating system. The lexicon pro- duced by this system is based on lexi- cal-semantic relations, but is compatible with a variety of other models of lexicon structure. The lexicon builder is suit- able for the generation of moderate-sized vocabularies and has been used to construct a lexicon for a small medical expert system. A future version of the lexicon builder will create a much larger lexicon by parsing definitions from machine-readable dictionaries. INTRODUCTION Natural language processing systems need much larger lexicons than those available today. Furthermore, a good com- puter lexicon with semantic as well as syntactic information is elaborate and hard to construct. We have created a program which enables its user to inter- actively build and extend a lexicon. The program sets up a user environment similar to a simple interactive operating system; in this environment lexical entries can be produced through a small set of commands, combined with prompts specified by the user for the desired kind of lexicon. The interactive lexicon builder is being used to help construct entries for a lexicon to be used to parse and generate stroke case reports. Many terms in this medical sublanguage either do not appear in standard dictionaries or are used in the sublanguage with special meanings. The design of the lexicon builder is inuended to be general enough to make it useful for others building lexicons for large natural language processing systems involving different sublanguages. The interactive lexicon builder will be the basis for a fully automatic lexicon builder which uses Sager's Linguistic String Parser (LSP) to parse machine- readable text into a relational network based on a modified version of Werner's NTQ (Modification-Taxonomy-Queueing) schema. Initially this program will be applied to Webster's Seventh Collegiate Dictionary and the Longman Dictionary of Contemporary English, both of which are available in machine-readable form. LEXICAL-SENANTIC RELATIONS The semantic component of the lexicon produced by this system consists princi- pally of a network of lexical-semantic relations. That is, the meaning of a word in the lexicon is indicated as far as possible by its relationships with other words. These relations often have seman- tic content themselves and thus contribute to the definition of the words they link. The two most familiar such relations are synonymy and antonymy, but others are interesting and important. For instance, to take an example from the vocabulary of stroke reports, the carotid is a kind of artery and an artery is a kind of blood Vessel. This "is a kind of" relation is taxonomy. We express the taxonomic rela- tions of "carotid', "artery" and "blood vessel" with the relational arcs carotid T artery artery T blood vessel Another important relation is …

Content truncated to 500 words