Domain-specific information retrieval using rcommenders

Abstract

The continuing increase in the volume of information available in our daily lives is creating ever greater challenges for people to find personally useful information. One approach used to addressing this problem is Personalized Information Retrieval (PIR). PIR systems collect a user's personal information from both implicit and explicit sources to build a user profile with the objective of giving retrieval results which better meet their individual user information needs than a standard Information Retrieval (IR) system. However, in many situations there may be no opportunity to learn about the specific interests of a user and build a personal model when this user is querying on a new topic, e.g. when a user visits a museum or exhibition which is unrelated to their normal search interests. Under this condition, the experiences and behaviours of other previous users, who have made similar queries, could be used to build a model of user behavior in this domain. My PhD proposes to focus on the development of new and innovative methods of domain-specific IR. My work seeks to combine recommender algorithms trained using previous search behaviours from different searchers with a standard ranked IR method to form a domain-specific IR model to improve the search effectiveness for a user entering a query without personal prior search history on this topic. The challenges for my work are: how to provide users better results; how to train and evaluate the methods proposed in my work.