AbstractsComputer Science

Abstract

This thesis combines aspects from two approaches to information access, information filtering and information retrieval, in an effort to improve the signal to noise ratio in interfaces to conversational data. These two ideas are blended into one system by augmenting a search engine indexing Usenet messages with concepts and ideas from recommender systems theory. My aim is to achieve a situation where the overall result relevance is improved by exploiting the qualities of both approaches. Important issues in this context are obtaining ratings, evaluating relevance rankings and the application of useful user profiles. An architecture called NewsView has been designed as part of the work on this thesis. NewsView describes a framework for interfaces to Usenet with information retrieval and information filtering concepts built into it, as well as extensive navigational possibilities within the data. My aim with this framework is to provide a testbed for user interface, information filtering and information retrieval issues, and, most importantly, combinations of the three.