Fisheye Matching: viewpoint-sensitive feature generation based on concept structure

Recent rapid growth of information environment such as the Internet makes i t easy for us to get vast information. On the other hand, "information over flow" is becoming a serious problem. To cope with such a problem, we have e xtended the normal Vector Space Model (VSM) to reflect the users' viewpoint s more clearly. We call this new matching method the Fisheye Matching metho d, which generates the features related to the users' viewpoints based on t he concept structure of an electronic dictionary. In the Fisheye Matching m ethod, the users' viewpoints are expressed as a set of word groups, each of which corresponds to a certain concept in the concept structure. Each conc ept in the dictionary has heading information, and the users can grasp thei r viewpoints easily from such information. Experimental results on informat ion retrieval show that the Fisheye Matching method can not only retrieve d ocuments in which the users take interest, but also supply them with useful information on their viewpoints. (C) 2000 Elsevier Science B.V. All rights reserved.