With the rapid expansion of the Internet, information retrieval tasks are increasingly performed by casual users on Web search engines without the help of human intermediaries. From a retrieval perspective, the Web is a vast heterogeneous database covering a large variety of topics and different depths. This poses a challenge to the performance of Web retrieval systems in the absence of search intermediaries. It has been argued forcefully that exploiting the user's context has the potential to improve Web retrieval systems as more information is available about a user and his/her information need. A retrieval system is said to have the ability of "personalization" if it is able to incorporate the context of a user's search in the adaptation to the user's specific information need. In this talk, I will present our approach of personalization that is based on the role-based context model. Acting as an experienced intermediary, the retrieval system incorporates the possible context information by inferring the role of the user's search. This approach enables relevant context information to come from both the user's previous search history and that of others who share the same role. I will also talk about some interesting issues we have encountered during the process of constructing the model.
For the colloquium series schedule, see the UMD Computational Linguistics Colloquium Series web page at http://umiacs.umd.edu/~resnik/cl_colloquium/. If you are interested in meeting with the speaker, please contact Philip Resnik (resnik@umiacs.umd.edu).