MUSIC AMBIGUITY GENE: CRITERIA FOR EMOTION-BASED TOP-N MUSIC RECOMMENDATION

摘要

An emotion-based music recommender system faces the challenge to match every user's expectation perfectly due to the subjective nature of human perception. In this paper, Music Ambiguity Gene (MAG) is defined as an indicator of human perception diversity on the emotion of a piece of music. A decision tree algorithm is used to build a model which predicts MAG from a set of musical features. An approach to provide Top-N recommended music items according to the value of MAG of each item is proposed based on the fact that MAG could have an impact on users' rating of recommended items. Experimental results demonstrate the effect of using MAG as criteria to give Top-N recommendation.