Output details
11 - Computer Science and Informatics
Robert Gordon University
Finding the Hidden Gems: Recommending Untagged Music
This paper was accepted for oral presentation at this world-leading AI conference (17.1% acceptance [IJCAI Preface]). Although the paper’s topic is music recommendation, its significance comes from its evaluation method that uses social media as a proxy for a user trial. Its novel Association Score combines statistical data on listening and liking tracks available in last.fm to estimate the quality of recommendations. Subsequently, Association Score has been compared with popular proxies (genre, year, cultural similarity) against a real user study for music recommendation; association score shows the best correlation with relevance from real users [Horsburgh’s PhD Thesis http://hdl.handle.net/10059/859].