An Object-Oriented and Fast Lexicon for Semantic Generation

TL;DR

This paper presents a novel object-oriented approach for efficiently storing and accessing a large computational lexicon in Prolog, outperforming traditional relational database methods for semantic generation tasks.

Contribution

It introduces an object-oriented storage model for a detailed HPSG-style lexicon, demonstrating efficient retrieval in Prolog, which is more suitable than relational models.

Findings

Object-oriented storage improves access speed for large lexicons.

Indexing and compression techniques enable efficient retrieval in Prolog.

Object-oriented model fits better with detailed lexical specifications.

Abstract

This paper is about the technical design of a large computational lexicon, its storage, and its access from a Prolog environment. Traditionally, efficient access and storage of data structures is implemented by a relational database management system. In Delilah, a lexicon-based NLP system, efficient access to the lexicon by the semantic generator is vital. We show that our highly detailed HPSG-style lexical specifications do not fit well in the Relational Model, and that they cannot be efficiently retrieved. We argue that they fit more naturally in the Object-Oriented Model. Although storage of objects is redundant, we claim that efficient access is still possible by applying indexing, and compression techniques from the Relational Model to the Object-Oriented Model. We demonstrate that it is possible to implement object-oriented storage and fast access in ISO Prolog.