Goal-directed complete-web recommendation

• Authors: Tingshao Zhu
• Venue: Goal-directed complete-web recommendation
• Year: 2006

While the World Wide Web (WWW) contains a vast quantity of information, it is often difficult for Web users to find the information they seek. There are many recommender systems that are designed to help users find relevant information on the Web; however, as many of these systems are server-side, they can only provide information about one specific Web site and they are typically based only on correlations amongst the pages that the various users visit. Unfortunately, there is no reason to believe that these correlated pages will necessarily contain useful information. Here, a passive Goal-Directed Complete-Web (GCW) recommender system, which recommends relevant pages from anywhere on the Web to satisfy the user's current information need without any explicit additional input, has been developed. After identifying the search strategy that is employed by actual users while they browse the Web, the model attempts to locate the pages that satisfy the user's information need based on the content of the pages the user has visited, and the actions the user has applied to these pages. To build such models, I develop a number of browsing features ---browsing properties of the words, in the context of the current session---to capture the actions of the Web user. Because the method is based on how the words are used (while training on these browsing feature values), it can be applied to make predictions about pages that have never been visited. This model is therefore independent of users, specific words and specific Web pages, and so it can be used to identify relevant pages in any new Web environment. To evaluate the predictive models, we have conducted two user studies, each involving over one hundred participants. Data from the user studies demonstrate that the models can effectively identify the information needs of new users, leading them to previously unseen, but relevant pages.