Content-based music recommendation using underlying music preference structure

Abstract

The cold start problem for new users or items is a great challenge for recommender systems. New items can be positioned within the existing items using a similarity metric to estimate their ratings. However, the calculation of similarity varies by domain and available resources. In this paper, we propose a content-based music recommender system which is based on a set of attributes derived from psychological studies of music preference. These five attributes, namely, Mellow, Unpretentious, Sophisticated, Intense and Contemporary (MUSIC), better describe the underlying factors of music preference compared to music genre. Using 249 songs and hundreds of ratings and attribute scores, we first develop an acoustic content-based attribute detection using auditory modulation features and a regression by sparse representation. We then use the estimated attributes in a cold start recommendation scenario. The proposed content-based recommendation significantly outperforms genre-based and user-based recommendation based on the root-mean-square error. The results demonstrate the effectiveness of these attributes in music preference estimation. Such methods will increase the chance of less popular but interesting songs in the long tail to be listened to.

Leveraging the Structure of Musical Preference in Content-Aware Music Recommendation

“…The hyperparameters λW and λH are tuned on the validation set using the normalized discounted cumulative gain (NDCG, see below), and λB = 10 −2 . Since content-free WMF cannot perform an out-of-matrix recommendation task, we implemented a pure content-based method as a baseline for cold-start recommendation based on [23]. This method consists in computing a mean AVD preference vector for each user from the training set, and then performing recommendations based on similarities between this mean AVD preference vector and the AVD factors extracted from novel songs.…”

“…Recent studies [19,20,21] notably show that musical preference can be described by using a set of three factors termed arousal, valence and depth (AVD). The usefulness of musical preference models for recommendation has been pointed out in [22], but to the best of our knowledge, it has only been exploited in [23]. However, this approach is not based on collaborative filtering and relies on expert ratings for estimating a genre-specific musical preference model, which hinders its deployment at a larger scale.…”

A Deep Temporal Neural Music Recommendation Model Utilizing Music and User Metadata

“…Bogdanov and Herrera [17] evaluate the usage of metadata information in content based recommendation systems. Soleymani et al [18] compute the similarities between music based on their Mellow, Unpretentious, Sophisticated, Intense and Contemporary attributes. These attributes are extracted from the auditory modulation features of music.…”

Language independent recommender agent

“…Apart from IR techniques, Bayesian classifiers and different machine learning techniques like clustering decision trees and artificial neural networks can be used for the similarity calculation (Adomavicius & Tuzhilin, 2005). Content-based recommendation systems have been used by the researchers on various topics (Lops et al ., 2011), such as music (Soleymani et al ., 2015), books (Mooney & Roy, 2000), movies (Debnath et al ., 2008), etc. However, these systems need the items to be marked with features, which is usually objective, and cannot make use of user similarities.…”