Montague grammars as deductive databases

Abstract

This paper presents a technique for further applications of deductive databases in the field of natural language processing. Although there were enormous research efforts on analyzing the semantics of natural languages (NL), especially using Montague grammars (PTQ) for Prolog, corresponding programs for deductive databases are still missing, since range restriction is required in nearly all deductive database systems. But without a way to handle open facts and lambda expressions within logic programs, Montague grammars do not work. On the other side, several NL-syntax parsers, based on DCG grammars, using a deductive database, have worked very efficiently in daily use for several years.

This paper fills the gap and presents a technique for (simplified) Montague grammars, including a lambda-reduction, for deductive database systems. Since on the one hand deductive databases have several well known advantages over Prolog, such as arbitrary recursion, integrated database access for the dictionary, easy handling of a large amount of data to be analyzed, and since on the other hand morphosyntactical and syntactical analysis of natural language texts already work very efficiently, there is a great need to make semantic analysis via Montague grammars available for bottom-up and set-oriented deductive databases as well.

We describe an implementation of the PTQ framework, including a translator and logic simplifier, based on deductive database systems. We have fully implemented the material presented in the deductive database system LOLA.