Oramas, S., Espinosa-Anke L., Lawlor A., Serra X., & Saggion H. Exploring Customer Reviews for Music Genre Classification and Evolutionary Studies. 17th International Society for Music Information Retrieval Conference (ISMIR'16)

In this paper, we explore a large multimodal dataset of about 65k albums constructed from a combination of Amazon customer reviews, MusicBrainz metadata and AcousticBrainz audio descriptors. Review texts are further enriched with named entity disambiguation along with polarity information derived from an aspect-based sentiment analysis framework. This dataset constitutes the cornerstone of two main contributions: First, we perform experiments on music genre classification, exploring a variety of feature types, including semantic, sentimental and acoustic features. These experiments show that modeling semantic information contributes to outperforming strong bag-of-words baselines. Second, we provide a diachronic study of the criticism of music genres via a quantitative analysis of the polarity associated to musical aspects over time. Our analysis hints at a potential correlation between key cultural and geopolitical events and the language and evolving sentiments found in music reviews.

Additional material:

• Preprint version

• MARD: Multimodal Album Reviews Dataset (details and download)

• Subset of MARD used for genre classification is released together with the evaluation code in the following GitHub repository

• The MARD dataset will be introduced in the next ISMIR tutorial "Natural Language Processing for MIR" https://wp.nyu.edu/ismir2016/event/tutorials/