Representative-Based Diversity Retrieval

Case-based reasoning has been widely used in electronic commerce by providing product recommendations that are most similar to user´s needs. However in some cases, the most similar recommendations are not necessary the ones most acceptable or useful. There is a growing awareness of the need for providing alternatives to similarity-based retrieval, alternatives that attempt to improve the recommendation diversity. In this paper, we present a new retrieval algorithm which is able to optimize the trade-off between similarity and diversity. Experimental results show that the proposed algorithm can achieve a better overall quality than other approaches.