Precision and recall of ranking information-filtering systems

摘要

An information-filtering system collects incoming data of specific interests described in a server profile. These systems can export their collections by submitting their profiles to a directory server, where users can query for relevant systems to answer their requests. We develop a new similarity measure to rank information-filtering systems for Boolean queries. Users can send queries to the top-ranked systems and obtain most of the relevant information. In contrast to an existing method developed by Radecki, our method requires less time and space complexity and has better recall and precision for higher-ranked systems.