Wei Li - PhD Transfer Talk - 31st May 2012

Video Category: 
Transfer Talk

Title: Using Community Trained Recommender Models for Enhanced Information Retrieval

Abstract: My research focuses on exploiting recommender models for enhanced information retrieval (IR). In recent years personalization has become a significant research area in seeking to assist users in finding information which is relevant to their current information needs. Personalization of IR systems operates by incorporating a model of the user's interests or personal information. However, in some situations, personal information is not available meaning that such personalization techniques cannot be applied. In order to address this challenge, my research proposes a novel IR model which combines a recommender model with traditional IR methods to improve retrieval results for search tasks where the IR system has no opportunity to learn prior information about the user's knowledge of a domain for which they have not searched before. My models exploits search behaviour data from other previous users to build topic model groups which contain several different topic models based on different topic interests. When a user enters a query on a topic which is new to this user, appropriate topic groups are selected and used to predict ranking which this user may find interesting based on the behaviour of previous users who have similar topic interests. The recommender outputs are used in combination with the output of a standard IR system to produce the overall output to the user. Initial experimental results show that the model can improve retrieval effectiveness significantly.