Details

Autor(en) / Beteiligte

• Castillo, Jaime Ramírez
• Flores, M. Julia
• Leray, Philippe

Titel

Predicting spotify audio features from Last.fm tags

Ist Teil von

• Multimedia tools and applications, 2024-05, Vol.83 (16), p.48311-48330

Ort / Verlag

New York: Springer US

Erscheinungsjahr

2024

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Music information retrieval (MIR) is an interdisciplinary research field that focuses on the extraction, processing, and knowledge discovery of information contained in music. While previous studies have utilized Spotify audio features and Last.fm tags as input values for classification tasks, such as music genre recognition, their potential as target values has remained unexplored. In this article, we address this notable gap in the research landscape by proposing a novel approach to predict Spotify audio features based on a set of Last.fm tags. By predicting audio features, we aim to explore the relationship between subjective perception and concrete musical features, shedding light on patterns and hidden correlations between how music is perceived, consumed, and discovered. Additionally, the predicted audio features can be leveraged in recommendation systems to provide users with explainable recommendations, bridging the gap between algorithmic suggestions and user understanding. Our experiments involve training models such as GPT-2, XGBRegressor, and Bayesian Ridge regressor to predict Spotify audio features from Last.fm tags. Through our findings, we contribute to the advancement of MIR research by demonstrating the potential of Last.fm tags as target values and paving the way for future research on the connection between subjective and objective music characterization. Our approach holds promise for both listeners and researchers, offering new insights into the intricate relationship between perception and audio signal in music. Our study aims to explore the feasibility and efficacy of this unique approach, where we intentionally refrain from using traditional audio-based or metadata-driven methods.