Ways of Computing Diverse Collaborative Recommendations

Abstract. Conversational recommender systems adapt the sets of products they recommend in light of user feedback. Our contribution here is to devise and compare four different mechanisms for enhancing the diversity of the recommendations made by collaborative recommenders. Significantly, we increase diversity using collaborative data only. We find that measuring the distance between products using Hamming Distance is more effective than using Inverse Pearson Correlation.

• Type: Conference
• Year: 2006
• Where: AH
• Authors: Derek G. Bridge, John Paul Kelly
• Keywords: AH 2006, Collaborative Recommenders, Conversational Recommender Systems, Inverse Pearson Correlation