A Semi-supervised Topic-based User Model for Web Information Visualization

Saleheen, S. and Lai, W.

    Web graph is a very effective tool to visualize web information. To serve user specific visualization and to reduce the size, personalization is applied to web graphs, by integrating user interests in filtering, graph generation and clustering. Modelling of the user interests is the key aspect to achieve personalization. Keyword-based user models are being used frequently for faster results. However, shortcomings of keywordbased models include lack of semantics that leads to inefficient similarity measurement and categorization. An ontological user profile can solve the polysemy problem inherent in a keyword-based profile. Because ontological profiles are mostly domain specific, they cannot be used efficiently in general web space which has information from diverse domains. This paper presents a topic-based hierarchical user model to address cross-domain interest-based visualization. The development and update procedures of the model consult theWordNet and/or the user. In connection with that, user interest-based measures are provided for graph generation, such as term and document similarity as well as document relatedness. These play an important role in user interest-based filtering and clustering. An experiment shows the effectiveness of the model over its keyword-based counterpart.
Cite as: Saleheen, S. and Lai, W. (2015). A Semi-supervised Topic-based User Model for Web Information Visualization. In Proc. 11th Asia-Pacific Conference on Conceptual Modelling (APCCM 2015) Sydney, Australia. CRPIT, 165. Saeki, M. and Kohler, H. Eds., ACS. 43-52
pdf (from crpit.com) pdf (local if available) BibTeX EndNote GS