README.md

---
dataset_info:
- config_name: default
  features:
  - name: id
    dtype: string
  - name: type
    dtype: string
  - name: full_audio
    dtype: audio
  - name: background_audio_quality
    dtype: float64
  - name: overall_audio_quality
    dtype: float64
  splits:
  - name: train
    num_examples: 71638
- config_name: clean
  features:
  - name: id
    dtype: string
  - name: type
    dtype: string
  - name: full_audio
    dtype: audio
  - name: background_audio_quality
    dtype: float64
  - name: overall_audio_quality
    dtype: float64
  splits:
  - name: train
    num_examples: 47944
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
- config_name: clean
  data_files:
  - split: train
    path: clean/train-*
---

Merged audio dataset with `background_audio_quality` / `overall_audio_quality` scores.

- **default**: all 71638 rows.
- **clean**: quality-filtered subset (background_audio_quality >= 2.5 AND overall_audio_quality >= 2.2),
  with `type` relabeled (literotica / pornhub / audio_love). 47944 rows.