2004 年 11 巻 3 号 p. 252-261
In this paper, the technique for extracting user's interests implicit in E-mails and recommending with surprisingness by presenting unexpected contents is proposed. We discover two effective measures for extracting such interests through some experiments. One is the number of the times a user sent for selecting interesting Topics and the other is the modified tf · idf value for selecting interesting words. A Web page recommendation system which applies this technique is also introduced. This system can guide a user ahead towards unexpected contents which are not limited in the specific domain.