ファジィ集合理論に基づく重み付き文献検索システム

細野公男; Kimio Hosono; 高柳敏子; Toshiko; 後藤智範; Tomonori Gotoh; 原田隆史; Takashi Harada

doi:10.46895/lis.23.137

ファジィ集合理論に基づく重み付き文献検索システムWeighted document retrieval systems based on fuzzy set theory approaches

細野公男¹，高柳敏子²，後藤智範³，原田隆史³Kimio Hosono¹, Toshiko ², Tomonori Gotoh³, Takashi Harada³

¹ 慶應義塾大学文学部図書館・情報学科教授Professor, School of Library and Information Science, Keio University ◇ 〒108-8345 東京都港区三田二丁目15番45号 ◇ Mita 2-15-45, Minato-ku, Tokyo 108-8345, Japan

² 獨協大学経済学部助教授Associate Professor, Faculty of Economics, Dokkyo University ◇ 〒340-0042 埼玉県草加市学園町1番1号 ◇ Gakuencho 1-1, Soka-shi, Saitama 340-0042, Japan

³ 慶應義塾大学大学院文学研究科図書館・情報学専攻Graduate School of Library and Information Science, Keio University ◇ 〒108-8345 東京都港区三田二丁目15番45号 ◇ Mita 2-15-45, Minato-ku, Tokyo 108-8345, Japan

発行日：1986年3月25日Published: March 25, 1986

There exist unavoidable drawbacks in the conventional document retrieval systems based on Boolean logic. Every keyword assigned to the documents in the collection is given equal weight irrespective of its actual relatedness to each document. This aspect is also applicable to the search formula where keywords are treated equally, and there is no chance for a searcher to specify the extent of the keyword’s significance. In addition, AND operators are excessively influential when narrowing the range of the documents to be retrieved.

In order to get rid of the defects, a variety of techniques have been introduced into retrieval systems such as models using conditional probability theory, vector space, and so forth. Fuzzy set theory approaches are among these and are quite promising for this purpose since human behavior is essentially so fuzzy that it is difficult to be represented by, for example, probability terms.

This paper describes following items concerning fuzzy retrieval systems from this viewpoint. 1) General characteristics of fuzzy retrieval. 2) Roles and functions of the weight. 3) Calculation formula for the ‘Retrieval Status Value’. 4) Manipulating methods of the metamorphic functions of Boolean logical operators.

Finally, this paper mentions some problems of determining legitimate RSV and e₁ formula, especially for handling NOT function, and of selecting adequate Membership Functions, when developing operational fuzzy retrieval systems.