Creating and Querying Lexical Data Bases

CREATING AND QUERYING LEXICAL DATA BASES Mary S. Neff, Roy J. Byrd, and Omneya A. Rizk IBM T. J. Watson Research Center P. O. Box 704 Yorktown Heights, New York 10598 ABSTRACT Users of computerized dictionaries require powerful and flexible tools for analyzing and manipulating the informa- tion in them. This paper discusses a system for grammat- ically describing and parsing entries from machine-readable dictionary tapes and a lexicai data base representation for storing the dictionary information. It also describes a lan- guage for querying, formatting, and maintaining diction- aries and other lexical data stored with that representation. 1. Introduction. Computer resident lexical information is required in a large variety of applications, including on-line dictionary refer- ence systems (Neff and Byrd(1987)), lexicographic sys- tems for creating new dictionaries (Tompa(1986)), systems for performing lexicological analyses of existing dictionaries (Zampolli and Calzolari(1985), Chodorow et a1.(1985)), and various natural language processing sys- tems (Byrd(1986)). Establishment of standard represent- ations and access mechanisms for lexical information will help us to better meet these requirements. The Lexical Systems project at the IBM T. J. Watson Re- search Center has defined a lexical data base representation (LDB) for lexical information, a set of tools and grammars for converting machine readable dictionary (MRD) type- setting tapes to the LDB representation, and a Lexical Query Language (LQL) for specifying storage and retrieval operations on LDBs. This paper briefly discusses the LDB structure, the tools and grammar for creating LDBs from existing sources, and then describes the facilities and im- plementation of LQL. Of the many machine-readable dictionary (MRD) re- sources available, current research concerns exploitation of the information from one dictionary (for example, Michiels (1982), Ahlswede, et. al. (1986)) or, in the case of multiple dictionaries, a limited subset of information from each (cf. Chodorow, et. al. (1985)). This exploitation is usually via a special extraction program which gets only the information needed by the project or application. Al- though in some cases the literature mentions the term "data base", the term is frequently not used in any strict sense, as it is here. The work reported in the present paper is related to work reported by Calzolari (1984a, 1984b) and Calzolari and Picchi(1986). Some of it, particularly the development of tools and grammars for parsing entries in MRD tapes to LDB format, is most related in spirit to some of the work reported by the New Oxford English Dictionary project (Tompa (1986a), Gonnet and Tompa(1986), Benbow(1986)). Whereas the new OED project has de- fined a markup language for dictionary entries before data entry and a grammar based on that markup language, this work attempts to parse and convert to a standard form a number of different raw type-setting tapes without any pre-editing. 2. Lexicai Data Bases. Dictionary entries are typically organized as shallow hier- archies of information with a variable number of instances of certain items at each level, e.g. multiple homographs within an entry or multiple senses within a homograph. More formally, they can be characterized …

Content truncated to 500 words