REASON: NLP-based Search System for the WWW

Natural Language Processing for the World Wide Web

Volume

Papers from the 1997 AAAI Spring Symposium

Contents

This paper presents REASON, an NLP system designed for knowledge-based information search in large digital libraries and the WWW. Unlike conventional information retrieval methods, our approach is based on using the content of natural language texts. We have elaborated a searching strategy consisting of three steps: (1) analysis natural language documents and queries resulting in adequate understanding of their content; (2) loading the extracted information into the knowledge base and in this way forming its semantic representation; (3) query matching proper consisting in matching the content of the query against the content of the input documents with the aim of finding relevant documents. The advanced searching capabilities of REASON are provided by its main constituents: the Language Model, the Knowledge Base, the Dictionary and the Software specially developed for content understanding and content-based information search. We characterize the particularities of the System and describe the basic components of the formal linguistic model, the architecture of the knowledge base, the set-up of the Dictionary, and the main modules of the System - the module of analysis and the logical inference module. REASON is an important contribution to the development of efficient tools for information search.