lossminimilization/Music_by_Emotion

---
license: other
license_name: hi
license_link: LICENSE
task_categories:
- audio-classification
language:
- en
tags:
- music
- emotion
- recognition
- classification
pretty_name: Music by Emotion
---
🎵 Music by Emotion Dataset
Dataset Summary

The Music by Emotion dataset is a custom audio dataset designed for supervised music emotion recognition tasks. It consists of 1,000 music samples, each a 30-second audio clip, sourced from publicly available SoundCloud content.

Each clip is labeled according to its emotional content, derived from musical characteristics such as tempo and musical key.

🎯 Task

Task type: Audio Classification

Domain: Music Emotion Recognition

Input: 30-second music audio clips

Output: Emotion labels

📂 Dataset Composition

Total samples: 1,000

Clip duration: 30 seconds

Audio source: Publicly available SoundCloud tracks

Format: Audio files suitable for machine learning pipelines

🏷️ Labeling Methodology

Emotion labels were assigned by analyzing two primary musical features:

Tempo:

Slow

Medium

Fast

Musical key:

Major

Minor

These musical attributes were used to infer the emotional characteristics of each clip.

🎭 Emotion Classes

The dataset includes the following 11 label categories:

Label Description
Angry High-intensity, aggressive musical characteristics
Contempt Dismissive or disdainful emotional tone
Disgust Harsh or unpleasant musical qualities
Fear Tense or suspenseful musical patterns
Happy Upbeat, positive emotional tone
Neutral Emotionally balanced or ambiguous content
Sad Slow, minor-key, melancholic music
Sleepy Low-energy, calming, or drowsy music
Surprised Sudden changes or unexpected musical elements
Bad Poor audio quality
Boring Incorrect or irrelevant format (e.g., podcasts, spoken content)

Note:

The Bad label indicates low-quality audio that may not be suitable for training models.

The Boring label identifies non-music or incorrectly formatted content, such as podcasts.

🔍 Intended Uses

Music emotion classification

Audio representation learning

Benchmarking emotion recognition models

Educational and research applications in music information retrieval (MIR)

⚠️ Limitations

Emotion labeling is inherently subjective.

The dataset size (1,000 samples) may limit generalization for large-scale models.

Labels are inferred from musical features rather than listener studies.

Some samples are intentionally labeled as Bad or Boring, which may require filtering before training.

📜 Licensing & Source Information

Audio source: Publicly available SoundCloud content

License: Users are responsible for ensuring compliance with SoundCloud’s terms of service and applicable licenses before redistribution or commercial use.

🙌 Acknowledgements

SoundCloud artists for publicly shared audio content

Hugging Face for dataset hosting and tooling

📬 Contact / Contributions

Contributions, corrections, and improvements to labeling or metadata are welcome via pull requests.