A generalised approach to similarity-based retrieval in recommender systems

Recommender systems for helping users to select from available products or services are increasingly common in electronic commerce. Typically in case-based reasoning (CBR) approaches to product recommendation, the items recommended are those that are most similar to a target query representing the elicited requirements of the user. Usually in practice, the user is required to specify a single preferred value for each attribute in the query. However, we argue that a more flexible approach to requirements elicitation is necessary to meet the needs of different users, ranging from those who know exactly what they are looking for to those whose requirements are vague in the extreme. We show how the standard approach to similarity-based retrieval can be generalised to support queries in which the user can enter any number of preferred values of a selected attribute, and examine the potential benefits of the approach.