|
|
--- |
|
|
language: en |
|
|
tags: |
|
|
- audio |
|
|
- speech-recognition |
|
|
- wav2vec2 |
|
|
- tokenizer |
|
|
license: apache-2.0 |
|
|
--- |
|
|
|
|
|
# my-wav2vec2-processor |
|
|
|
|
|
This is a Wav2Vec2 processor (tokenizer + feature extractor) for speech recognition. |
|
|
|
|
|
- **Base model**: facebook/wav2vec2-base-960h |
|
|
- **Uploaded by**: Utkarshg02 |
|
|
- **Intended use**: Preprocessing audio data for Automatic Speech Recognition (ASR) tasks. |
|
|
|
|
|
## How to Use |
|
|
|
|
|
```python |
|
|
from transformers import Wav2Vec2Processor |
|
|
|
|
|
processor = Wav2Vec2Processor.from_pretrained("Utkarshg02/my-wav2vec2-processor") |
|
|
|
|
|
# Example: Processing an audio array |
|
|
inputs = processor(audio_array, sampling_rate=16000, return_tensors="pt") |
|
|
|
|
|
|
|
|
## Limitations and Biases |
|
|
- Only provides preprocessing (feature extraction + tokenization). |
|
|
- Base model is trained on English; may not work well on other languages or accents. |
|
|
|
|
|
## License |
|
|
|
|
|
This processor follows the same license as the base model: Apache 2.0. |
|
|
|
|
|
## References |
|
|
|
|
|
- [facebook/wav2vec2-base-960h](https://huggingface.co/facebook/wav2vec2-base-960h) |
|
|
- [Wav2Vec2 Paper](https://arxiv.org/abs/2006.11477) |
|
|
|
