thesven/whisper-tiny-bn-v2
Automatic Speech Recognition
•
Updated
•
4
•
1
input_features
sequence | labels
sequence |
|---|---|
[[-0.5007599592208862,-0.5007599592208862,-0.5007599592208862,-0.5007599592208862,-0.500759959220886(...TRUNCATED) | [50258,50302,50359,50363,29045,97,29045,123,29045,101,29045,123,220,29045,237,156,2250,29045,224,220(...TRUNCATED) |
[[-0.5822842121124268,-0.5822842121124268,-0.5822842121124268,-0.5822842121124268,-0.582284212112426(...TRUNCATED) | [50258,50302,50359,50363,29045,243,156,100,225,29045,97,156,100,235,29045,97,29045,123,156,2250,2904(...TRUNCATED) |
[[-0.8000154495239258,-0.8000154495239258,-0.8000154495239258,-0.8000154495239258,-0.521644949913024(...TRUNCATED) | [50258,50302,50359,50363,29045,97,29045,123,29045,101,29045,123,220,29045,97,29045,122,156,12811,220(...TRUNCATED) |
[[-0.5339429378509521,-0.5339429378509521,-0.5339429378509521,-0.5339429378509521,-0.533942937850952(...TRUNCATED) | [50258,50302,50359,50363,29045,97,29045,123,29045,101,29045,123,220,156,2250,29045,123,29045,250,290(...TRUNCATED) |
[[-0.5761864185333252,-0.5761864185333252,-0.5761864185333252,-0.5761864185333252,-0.576186418533325(...TRUNCATED) | [50258,50302,50359,50363,29045,237,29045,253,29045,123,220,29045,106,156,100,224,29045,110,29045,97,(...TRUNCATED) |
[[-0.5368735790252686,-0.5368735790252686,-0.5368735790252686,-0.5368735790252686,-0.536873579025268(...TRUNCATED) | [50258,50302,50359,50363,29045,116,29045,94,156,15773,29045,243,29045,253,29045,123,220,156,2250,290(...TRUNCATED) |
[[-0.6104068756103516,-0.6104068756103516,-0.6104068756103516,-0.6104068756103516,-0.610406875610351(...TRUNCATED) | [50258,50302,50359,50363,29045,106,29045,122,29045,251,156,100,229,12,29045,106,29045,100,156,100,23(...TRUNCATED) |
[[-0.6414562463760376,-0.6414562463760376,-0.6414562463760376,-0.6414562463760376,-0.641456246376037(...TRUNCATED) | [50258,50302,50359,50363,156,2250,29045,110,29045,243,156,100,229,220,29045,227,156,2250,29045,106,1(...TRUNCATED) |
[[-0.5835813283920288,-0.5835813283920288,-0.5835813283920288,-0.5835813283920288,-0.583581328392028(...TRUNCATED) | [50258,50302,50359,50363,29045,116,29045,122,29045,224,29045,116,156,100,235,29045,243,156,100,225,2(...TRUNCATED) |
[[-0.6166214942932129,-0.6166214942932129,-0.6166214942932129,-0.6166214942932129,-0.160181879997253(...TRUNCATED) | [50258,50302,50359,50363,29045,107,29045,98,29045,122,156,12811,156,11061,29045,97,29045,123,220,290(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
Dataset Card for "bengali-ai-train-set-tiny"
Whisper Model Information
- Model Homepage: openai/whisper-tiny on Hugging Face
- Model Paper: Robust Speech Recognition via Large-Scale Weak Supervision
Dataset Summary
This dataset is designed to help finetune the openai/whisper-tiny model with additional information in the Bengali language. It consists of an additional 11,000 labeled audio samples from the OOD-Speech dataset, specifically designed for out-of-distribution benchmarking in Bengali.
Supported Tasks and Leaderboards
The primary task supported by this dataset is automatic speech recognition (ASR) in the Bengali language, specifically for finetuning the openai/whisper-tiny model.
Languages
The dataset is in Bengali.
Dataset Structure
Data Instances
Each instance in the dataset consists of an audio sample in Bengali along with its corresponding transcription.
Data Fields
audio: The audio sample in Bengali.transcription: The corresponding transcription of the audio sample in Bengali.
Data Splits
The dataset is split into training and validation sets. The training set consists of 10,000 samples, and the validation set consists of 1,000 samples.
Additional Information
Dataset Curators
The dataset has been curated by "thesven".
Licensing Information
Licensing information for the OOD-Speech dataset can be found in the original paper.
Citation Information
@article{OOD-Speech2023, title={OOD-Speech: A Large Bengali Speech Recognition Dataset for Out-of-Distribution Benchmarking}, author={Authors of the OOD-Speech paper}, journal={arXiv preprint arXiv:2305.09688}, year={2023} }
- Downloads last month
- 0
Homepage:
OOD-Speech: A Large Bengali Speech Recognition Dataset for Out-of-Distribution Benchmarking
Paper:
OOD-Speech: A Large Bengali Speech Recognition Dataset for Out-of-Distribution Benchmarking
Size of downloaded dataset files:
1.67 GB
Size of the auto-converted Parquet files:
1.67 GB
Number of rows:
11,000