An Approach to Personalizing Web Search Results

The excess of information available on the Web has required users to make a greater effort to retrieve information relevant to their interests. Although conventional search engines are capable of returning good quality results in response to most queries, they are still unable to offer these results efficiently. This paper proposes an architecture for a web search personalization system that employs the latent semantic indexing technique, adapted for the web environment, in conjunction with a user model constructed implicitly by tracking user navigation in the documents resulting from the search. On the one hand, the latent semantic indexing technique allows the semantic relationship between websites to be identified, providing a better ordering of results. On the other hand, the user model makes it possible to identify a user's interests in the search and improves the ordering of the results offered according to those interests.