---
dataset_info:
  features:
  - name: caption_id
    dtype: int64
  - name: track_id
    dtype: int64
  - name: caption
    dtype: string
  - name: is_valid_subset
    dtype: bool
  - name: familiarity
    dtype: int64
  - name: artist_id
    dtype: int64
  - name: album_id
    dtype: int64
  - name: path
    dtype: audio
  - name: duration
    dtype: float64
  - name: __index_level_0__
    dtype: int64
  splits:
  - name: train
    num_bytes: 3543021978.0
    num_examples: 746
  download_size: 2591774479
  dataset_size: 3543021978.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This is a mirror to the example dataset "The Song Describer Dataset: a Corpus of Audio Captions for Music-and-Language Evaluation" paper by Manco et al.

Project page on Github: https://github.com/mulab-mir/song-describer-dataset

Dataset on Zenodoo: https://zenodo.org/records/10072001

Explore the dataset on your local machine:

```python 
import datasets
from renumics import spotlight

ds = datasets.load_dataset('renumics/song-describer-dataset')
spotlight.show(ds)
```



![image/png](https://cdn-uploads.huggingface.co/production/uploads/63dd29ffaf221a78fa4ec8d1/bcoTy3h_Q0uUUo-m880Iu.png)