My research on Web intelligence has been focused on mining the log file of a Web site for knowledge about the structure and link usage of the Web site, that is, how Web pages are linked to each other and how frequently links have been traversed by users. Such knowledge is then used to adapt the presentation of the Web site, make link prediction, and help ranking search results.
We developed a novel clustering algorithm, called PageCluster, for clustering conceptually related Web pages on the basis of their link similarity. We used Markov chain models for link prediction. We also developed a novel algorithm for ranking search results, called PageRate, which is an improvement to the PageRank algorithm in Google, the most widely used Web search engine. The research has led to a number of top-quality publications, including papers in ACM Transactions on Internet Technology, ACM conferences on hypertext, Springer lecture notes.
My research on intelligent interfaces is concerned with developing intelligent interface agents that help users to solve complex tasks using an interactive software application by being sensitive to a user’s knowledge, misconceptions, goals and plans. An intelligent interface agent does not act between the software application and the user. It instead allows the user to directly manipulate the software application on his/her own and observes over the shoulder of the user to understand the context of his/her actions, recognises his/her goals and plans while he/she is performing a task. It can automate tasks delegated by the user, guide use of a software application, correct and explain errors, and make recommendations and suggestions. Typical interactive software applications include Microsoft Windows applications, operating systems, interactive editors and Web-based shopping malls. This research was partly conducted in a joint project, funded by the National Natural Science Foundation of China.