Bimodal context classification for modeling web user’s information needs

Abstract

The fact that the World Wide Web is being used for several purposes also implies that there exist various contextual factors in Web spaces with which users interact every day and users may have various information quality factors to consider according to their current context. In this regards, it is important for Web recommendation services to recognize what quality factors should be considered in current context in order to enhance user satisfaction. In this thesis, Web contextual factors were collected with the aim of suggesting new concept of context aware Web recommendation service in which various non-considered aspects of Web context as well as semantic aspects of user’s target information are considered. It was also shown that it is necessary to classify Web contexts based on the information quality factors users consider in their minds when they choose websites or Web pages. As a preliminary study, the results of Web Activity system analysis showed that there are four main quality factors: credibility, recency, popularity, and relevance. From survey data analysis, it is shown that user tasks can be clustered into two context groups based on the quality factors that users consider. Finally, the results of client-sided log data analysis and performances of our proposed algorithm showed that it is possible to enable Web recommendation services to infer the current context group and to identify high interested content pages to build user profile. This result implies that context recognition as well as precise user profile construction is possible using the limited data that are collected at client side.