Lexical Knowledge Engineering: MikroKosmos Revisited

This paper will describe the MikroKosmos methodology for the knowledge engineering of a computational lexicon used for text analysis. To do so, the paper outlines the general requirements for a knowledge base to be used for NLP, followed by specific requirements for building the lexical knowledge source. To highlight the issue of efficiency and reusability, the paper will contrast knowledge engineering for text analysis using the syntactic realizations of predicate logic against knowledge engineering for text analysis using lexical constraint satisfaction in combination with a world-model, or ontology. The paper suggests that the most efficient approach to the representation of natural language for analysis is the ontological and lexical semantic modeling (Onyshkevych & Nirenburg, 1994). Lexical semantics makes constraints accessible centrally from the concepts rather than through assertions of logic. References are made in this paper to the structuring of the English generation lexicon for the MikroKosmos KBMT project, a knowledgebased machine translation system based on deepmeaning analysis and text generation, developed at New Mexico State University, Computing Research Laboratory.