Automatic Selection of Defining Vocabulary in an Explanatory Dictionary

Abstract

One of the problems in converting a conventional (human-oriented) explanatory dictionary into a semantic database intended for the use in automatic reasoning systems is that such a database should not contain any cycles in its definitions, while the traditional dictionaries usually contain them. The cycles can be eliminated by declaring some words “primitive” (having no definition) while all other words are defined in terms of these ones. A method for detecting the cycles in definitions and selecting a minimal (though not the smallest) defining vocabulary is presented. Different strategies for selecting the words for the defining vocabulary are discussed and experimental data for a real dictionary are presented.

Work done under partial support of CONACyT, CGEPI-IPN, COFAA-IPN, and SNI, Mexico. Implementation by Gabriela Rivera-Loza. We thank Graeme Hirst and Ted Pedersen for useful discussion.

This is a slightly simplified real example from the Anaya dictionary of Spanish: “abeja: insecto que segrega miel”; “miel: sustancia que producen las abejas.”