Toward a Personalized Approach for Combining Document Relevance Estimates

摘要: A large body of work in the information retrieval area has highlighted that relevance is a complex and challenging concept. The underlying complexity appears mainly from fact estimated by considering multiple dimensions most them are subjective since they user-dependent. While used dimension topicality, recent works risen particularly personalized have shown personal preferences contextual factors such as interests, location task peculiarities to be jointly considered order enhance computation document relevance. To answer this challenge, commonly approaches based on linear combination schemes rely basically non-realistic independency property dimensions. In paper, we propose novel fuzzy-based aggregation operator able capture user’s importance well about their interaction. Our approach empirically evaluated relies standard TREC suggestion dataset involving 635 users 50 contexts. results highlight accounting for individual differences toward interaction introduces significant improvement performance.