Dataset Viewer
Auto-converted to Parquet Duplicate
speaker_id
stringlengths
1
4
audio0
audioduration (s)
0.5
0.5
audio1
audioduration (s)
0.5
0.5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
6
6
6
6
End of preview. Expand in Data Studio

Accented Speech Fragments

The Accented Speech Fragments dataset is a large curated collection of short speech samples designed for speaker identification and contrastive speaker embedding (voice printing).

Each speaker in the dataset is uniquely identified by a speaker ID and is associated with multiple speech samples (2 x 0.5s samples per row; 16 rows per speaker, totally 16 x '2 samples' x '0.5 seconds' = 16 seconds of raw audio). This structure makes the dataset particularly suitable for training contrastive learning models, where the goal is to distinguish between different speakers based on their voice characteristics.

Usage

In order to load the dataset from the hub, you can use the datasets library:

ds = datasets.load_dataset(
    "ThBel/Accented-Speech-Fragments",
    split='train', # 'train' or 'test'
    streaming=True # (optional)
)

for item in ds:
    # Do something with data
    print(item['speaker_id']) # or item['audio0'] or item['audio1']

Alternatively you may clone the ThBel/Accented-Speech-Fragments repository, and load the underlying parquet files using pandas.read_parquet.

Statistics and Image Format

Statistic Value
#Train samples 68416
#Speakers 2138
Sampling rate 16kHz
Audio duration 0.5s (8000 samples)

License and Intended Use

All audio samples contained in the dataset are licensed under CC BY-NC-SA 2.0 as per the license of the underlying Speech Accent Archive and are therefore supplied for research purposes only. The dataset may under no circumstances be used commercially.

Citation

If you use this dataset in your research, please provide appropriate credits to:

@misc{Weinberger2015,
  author = {Steven Weinberger},
  year = {2015},
  title = {Speech Accent Archive},
  institution = {George Mason University},
  howpublished = {Retrieved from \url{http://accent.gmu.edu}}
}

Disclaimer

I am not the original creator of the Speech Accent Archive and hold no rights over the original content. This dataset is provided as-is for research purposes, and all credit goes to the original creators.

Downloads last month
14