A Generalised Approach to Similarity-Based Retrieval in Recommender Systems

摘要

Recommender systems for helping users to selectfrom available products or services areincreasingly common in electronic commerce. Typically in case-based reasoning (CBR)approaches to product recommendation, the itemsrecommended are those that are most similar toa target query representing the elicitedrequirements of the user. Usually in practice,the user is required to specify a singlepreferred value for each attribute in thequery. However, we argue that a more flexibleapproach to requirements elicitation isnecessary to meet the needs of different users,ranging from those who know exactly what theyare looking for to those whose requirements arevague in the extreme. We show how the standardapproach to similarity-based retrieval can begeneralised to support queries in which theuser can enter any number of preferred valuesof a selected attribute, and examine thepotential benefits of the approach.