Contextual factors greatly affect users’ preferences for music, so they can benefit music recommendation and music retrieval. However, how to acquire and utilize the contextual information is still facing challenges. This paper proposes a novel approach for context-aware music recommendation, which infers users’ preferences for music, and then recommends music pieces that fit their real-time requirements. Specifically, the proposed approach first learns the low dimensional representations of music pieces from users’ music listening sequences using neural network models. Based on the learned representations, it then infers and models users’ general and contextual preferences for music from users’ historical listening records. Finally, music pieces in accordance with user’s preferences are recommended to the target user. Extensive experiments are conducted on real world datasets to compare the proposed method with other state-of-the-art recommendation methods. The results demonstrate that the proposed method significantly outperforms those baselines, especially on sparse data.

中文翻译：

---

基于序列的上下文感知音乐推荐

上下文因素极大地影响了用户对音乐的偏爱，因此他们可以受益于音乐推荐和音乐检索。然而，如何获取和利用上下文信息仍然面临挑战。本文提出了一种新颖的上下文感知音乐推荐方法，该方法可以推断用户对音乐的偏爱，然后推荐适合其实时需求的音乐作品。具体而言，所提出的方法首先使用神经网络模型从用户的音乐收听序列中学习音乐作品的低维表示。然后，基于学习到的表示，可以根据用户的历史收听记录来推断和建模用户对音乐的一般偏好和上下文偏好。最后，根据用户的喜好将音乐作品推荐给目标用户。在现实世界的数据集上进行了广泛的实验，以将提出的方法与其他最新推荐方法进行比较。结果表明，所提出的方法明显优于那些基线，尤其是在稀疏数据上。