Book Retrieval System Based on Flexible Matching between Natural Language Queries and Tables of Contents

摘要

Recent information-oriented society needs high-performed information retrieval(IR) systems. Book retrieval system is one of the important application of IR systems. Information exploited by book retrieval systems can be classified as follows: 1. titles, 2. indexing words and keywords, 3. tables of contents, 4. full texts. Among these, 3. tables of contents are the most ideal both in quality and in quantity and there has been a study of book retrieval system utilizing the structure of tables of contents. This paper proposes a new book retrieval method of using the Vector Space Model and flexible matching between natural language queries and tables of contents in addition to using the structure of tables of contents. We report experimental evaluation of the system using 1211 books of Iwanami expository series and 202 queries.