Web page organization and visualization using generative topographic mapping: A pilot study
Xiaofeng Zhang, Chak-Man Lam, William K. Cheung · 2004
Automatic Web page organization and visualization is an effective way for foraging information in a Web structure. Web pages contain both text (content) and links (structure), implying that content and structure analysis techniques should be adopted and properly integrated. In this paper, we take the probabilistic model-based approach and extend a topographypreserving model known as Generative Topography Map (GTM). The extended GTM provides a principled way to integrate Web pages and hyperlinks and project them into a low-dimension latent space (2D in our case) for visualization. The proposed extension has been applied to an artificially created dataset and also the WebKB dataset for performance evaluation. Based on the preliminary results obtained, we propose several directions for future research.