QUIXOTE : A Framework for Linguistic Information Processing

Abstract

In natural language processing, a very complex flow of information must be dealt with. There are two remarkable features in the current paradigm of computational linguistics : one is the constraint-based formalism of grammar and the other is the situatedness of semantics. In order to realize these features, we developed a language QUIXOTE as a framework for representing and processing such linguistic information. QUIXOE is a hybrid language of deductive object-oriented database and constraint logic programming language. The new mechanism of QUIXOT is a combination of the object-oriented notion such as object identity and the concept of module that classifies a large knowledge base. In addition, its logical inference is extended to be able to make restricted abduction. We first introduce the basic structure of QUIXOTE and thereafter we show that the typed-feature structure of constraint-based grammar can be represented in QUIXOTE, using complex objects and constraints upon them. Next, we show how QUIXOTE can contribute to the description of situation-based semantics, regarding modules as situations, where an abductive reasoning system for hidden information and an ambiguity analyzer of noun phrase reference are given as examples.