RecForUS: A Recommender System for Uncertain Scores

Download PDF

Abstract

We present RecForUs, a recommender system designed to offer accurate music recommendations through a competition between participants and an algorithmic recommender. Our framework aims to demonstrate the intricate management of uncertain scores in a recommender system, catering to the specific objectives of users. The demonstration showcases our novel RankDist algorithm that efficiently computes rank probabilities for items with uncertain scores, enabling optimal selection of ranking semantics tailored to different user objectives without requiring exhaustive evaluation of all possible worlds. RecForUS is versatile, demonstrating the effectiveness of generating top- 𝐾 query results in multiple scenarios.