Cache adaptation of internet services and user modeling for improving ranking precision in third-party results

This paper proposes a meta-search algorithm, which adapts to the user browsing behavior during his search sessions. In parallel, the algorithm is capable of monitoring the ability of its third-party search services in terms of refreshing and updating the content of their indexes. The algorithm uses five well known web search services namely AltaVista, Google, Lycos, MSN, and Yahoo!. The assessments made verified the robustness of our proposal, since we measured a significant improvement in the precision of the unified third-party results, especially for the lower recall levels where the user explores the top-ranked information. In addition, through capture recapture experiments we observed that Google, MSN and Yahoo! managed to adapt more adequately in the incessant evolution of the web, since they achieved higher freshness rates, managing in parallel to provide more validated active results to the end users.