Context-Sensitive Complementary Information Retrieval for Text Stream

With constant advances in information technology, more and more information is available and users’ information needs are becoming more diverse. Most conventional information systems only attempt to provide information that meets users’ specific interests. In contrast, we are working on ways of discovering information from the viewpoints of both interest and necessity. For example, we are trying to discover complementary information that provides additional knowledge on the users’ topics of interest, not just information that is similar to the topic. In previous work, which was based on extracting topic structures from closed-caption data, we proposed methods of searching for information to complement TV program content; that is, to provide users with more detailed information or different viewpoints. In this paper, we focus on the features of text streams (closed-caption data, etc.) and propose a method for context-sensitive retrieval of complementary information. We modified our topic-structure model for content representation and consider the “context” of a text stream in searching for complementary information. The “context” of the text stream is considered to be a series of topic structures. Based on such kind of context, we propose methods of searching for complementary information for TV programs, including query-type selection, query modification, and computation of the degree of complementarity. The experiment results showed that, comparing to our previous methods, the context-sensitive method could provide more additional information and avoid information overlap.