Affiliations: Graduate School of Information Science and Technology,
Hokkaido University, West 9, North 14, Kita-ku, Sapporo, Hokkaido 060-0814,
Japan | Graduate School of Engineering Science, Osaka
University, 1-3 Machikaneyama, Toyonaka, Osaka 560-8531, Japan
Note: [] Corresponding author: Tetsuya Yoshida, Graduate School of
Information Science and Technology, Hokkaido University, West 9. North 14,
Kita-ku, Sapporo, Hokkaido 060-0814, Japan. Tel.: +81 11 706 7253; Fax: +81 11
706 7808; E-mail: [email protected]
Abstract: Recent development of the network such as the Internet makes it easy
to retrieve documents over the World Wide Web (WWW). Document retrieval with
search engines is convenient, especially to retrieve documents for a domain
with which a user is not so familiar, since the retrieved documents can be used
to get a gross overview of the domain as a preliminary investigation. However,
when inappropriate query terms are used, it often results in retrieving too
many documents and the retrieved documents have to be scanned by the user
him/herself to find out truly valuable ones. This paper proposes a document
retrieval support system to reduce such "dual" retrieval by suggesting terms
based on the relationship of relevant terms to the specified query terms.
Semantic relations of terms are dealt with using a thesaurus and relevant terms
for the terms are organized into a tree structure based on the relations
defined in the thesaurus. Since the value of retrieved documents is
intrinsically user-dependent, even when the same query terms are used,
appropriate terms to facilitate document retrieval can vary depending on users.
To realize personalization in term suggestion, the relationship of terms for
the user is learned from a user-supplied case so that personalized terms can be
suggested. A prototype system has been implemented and it is evaluated through
experiments. The results of experiments are encouraging and show that it is
worth following this path, especially for supporting users who are not yet
familiar with the target domain.
Keywords: Document retrieval support, term suggestion, personalization, term relationship
