Developing a large lexical database for information retrieval, parsing, and text generation systems

This paper shows that it is possible to construct a lexical database by combining material from a number of machine-readable sources. We discuss the kind of lexical information required for applications in information retrieval and in other natural language processing areas, such as database interfaces and automatic filing systems. We describe the organization of our lexical database, which is stored in an Oracle Relational Database Management System and the design of the tables that comprise the database. In addition to the traditional alphabetic listing, access is provided from roots to derived forms and from derived forms to roots, and also through lexical and semantic relations between words, so that the database functions as a thesaurus as well as a dictionary. The database is designed to be open-ended and self-defining. Every attribute of every table is defined in the database itself. The lexical database can easily be extended through an SQL forms interface that facilitates additions to the tables.