Learning subjective relevance to facilitate information access
James R. Chen, Nathalie Mathé · 1995
As the amount of available electronic information is dramatically increasing, the ability for rapid and effective access to information has become critical. Most traditional information access methods rely on measures of relevance based on information content. We propose a new approachwhich augments existing information access methods with subjective relevance learned from user feedback. We developed an adaptive system which helps users access information by employing learned knowledge about which documents are likely to be relevant, given the current user's information need and user profile. This system is based on a model, called a relevance network,which learns and generalizes relevance information in a rapid, cost-effective, and incremental manner. We present the design of the relevance network and results of experimental evaluation.