Choosing Internal Representations for Natural Language

Abstract

The previous sections have given some of the flavor of how formulas may be produced from sentences through logic grammars. Basically, we have to transform a sentence into a structure in some internal meaning-representation language (first-order logic, lambda-calculus, etc.). The particular language chosen will depend on its suitability to represent the meaning of sentences with respect to the intended use of the structures obtained. We may want to use them for creating, consulting or modifying a data base, for having dialogues with an expert system, for extracting information about a constructive world, as an intermediate form in the translation from one language to another, as an aid for teaching some linguistic theory to a student, etc. Here we restrict ourselves to one of the best studied uses: querying knowledge or information stored.