2002 Volume 14 Issue 5 Pages 472-481
Web information visualization method based on the metaphor of immune network model is proposed to handle the set of documents obtained from the Web, such as retrieval results and online news articles. In order to achieve the effective browsing as well as to grasp the topic distribution through the document set, both the keyword map (a space on which keywords are arranged based on the co-occurrence) and the document clustering are generated in synthetic manner. The proposed method is based on the keyword extraction with using the activation mechanism of the immune network model, while the activation state of the network is also useful as a metaphor to improve the readability of the keyword map. Experimental results show that the document clusters covering about 70% of the document set without overlapping can be generated by the proposed method. The questionnaires are also issued to evaluate the proposed method from the viewpoint of topic representation, and the proposed method can get better results than k-means clustering, for 2 of 3 data sets. Furthermore, it is also shown that the topic-sensitive keyword map can be constantly obtained by introducing the immune network metaphor.
