Affiliations: Federal University of Uberlândia, Av. João Naves de Ávila, 2121 – 1B148, Uberlândia, MG 38400, Brazil. E-mail: [email protected] | Microsoft Research, One Microsoft Way, Redmond, WA 98052, USA. E-mail: [email protected]
Note:  Corresponding author.
Abstract: The quality of results to Web search queries is substantially limited because of the cost and short processing times allowed at search engine's data center to retrieve relevant pages, augment ads, and present them to the end-user. We tackle such an issue by proposing a radically different system, where the search engine replies to a query with a large list of relevant URLs. Our client-side platform then proceeds to download the target HTML files only, parse them, understand their content, and present a summary to the user. Different types of summaries can be created by using plug-ins attached to the base platform. Each plug-in provides the required functionality according to the type of summary desired by the user. With this novel mechanism, our system offers increased computational power for post-processing search results and consequently improves and personalizes the user's search experience while maintaining constant workload at the search engine. We present prototype implementations of the proposed search assistant and an associated shopping plug-in capable of detecting whether a Web-page encapsulates a direct commercial offering. We also review measurements related to projected system performance and demonstrate its applicability to different scenarios.