Beyond Search, Microsoft 2009
- Dates: June 10–11, 2009
- Location: Microsoft Research
Building 99, Lecture Room
Redmond, Washington, United States
“The Beyond Search – Semantic Computing and Internet Economics 2009 Workshop gathers the winners of the Beyond Search Request For Proposals (RFP), which directly addresses the need for more large-scale data by making additional real-world, large-scale data available to academia, under a limited data license agreement.”
As one of the winners, I was supposed to present my work in the conference, but unfortunately I did not make it because my US visa was still pending. Luckily, my PhD student Tamas Jambor represented me at the conference. The research is about:
Goal-Driven Information Retrieval
Jun Wang and Tamas Jambor
University College London, United Kingdom
Today’s search technologies heavily rely on textual queries of users to identify their information needs. Yet, users have various information needs. To cope with this, this research proposes a goal-driven information retrieval framework, in which the retrieval processes and combination strategies are influenced by automatically learning users’ information goals (in other words, types of user needs or retrieval tasks) and by estimating the probability of relevance between user information needs and documents. The research intends to establish a sequential approach to learning relevance by applying recent advances in Bayesian machine learning by explicitly modeling users’ various information goals, and, as a result, having the final algorithm respond to different information goals accordingly by a weighted average of different retrieval models. By counting the dependency of information needs among users, we will correlate users by applying hierarchical Bayesian methods.
Some of the studies were conduced when I worked in Microsoft, Cambridge. Other collaborators include With Mike Taylor, Onno Zoeter, Tom Minka and Steve Robertson in Microsoft, Cambridge
The slides can be found here.