User Interest Modeling and Its Application for Question Recommendation in User-Interactive Question Answering Systems

In this project, we propose to study the modeling of user interest in User-Interactive Question Answering (UIQA) systems. A generative model, the Topic based User Interest (TUI) model, will be invented to capture the user interests with latent topic method. The TUI model extracts interests for each user based on the questions that he has asked, as well as the social links that he interacts with others. Large-scale evaluations for question recommendation application will be conducted on data collections from real-world UIQA system. The Latent Dirichlet Allocation (LDA) model and the bag-of-words (BOW) model will be used as baselines for comparison. We propose to conduct this research in three phases. The first phrase will focus on building the user interest model. Exploring on how to use the user interest model for ranking questions will be conducted in Phase 2. Finally, we will evaluate the propose model in Phase 3.