Multimodal Music Emotion Dataset

This repository contains the dataset introduced in the paper:

A Multimodal Deep Network for Music Emotion Recognition using Audio Chorus and Lyrics

Mohammad Ali Talaghat, Elham Parvinnia, Mahdi Mehrabi, Reza Boostani

Published in IEEE Access

---

Description

The dataset is designed for research in music emotion recognition (MER) based on the valence-arousal model.

---

Dataset Details

• spotify_id — Unique track identifier from Spotify
• deezer_song_id — Unique track identifier from Deezer
• msd_song_id — Million Song Dataset (MSD) song ID
• msd_track_id — Million Song Dataset (MSD) track ID
• valence — Emotion annotation (positivity measure)
• arousal — Emotion annotation (energy measure)
• artist — Artist name
• track — Track title
• dataset_split — Data split indicator (e.g., train/validation/test)

Downloads

• Music Audio (Chorus)
• Music Lyrics

---

Citation

@article{talaghat2025mer,
  title={A Multimodal Deep Network for Music Emotion Recognition using Audio Chorus and Lyrics},
  author={Talaghat, Mohammad Ali and Parvinnia, Elham and Mehrabi, Mahdi and Boostani, Reza},
  journal={IEEE Access},
  year={2025}
}