Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,138 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- thesven/bengali-ai-train-set-tiny
|
| 4 |
+
language:
|
| 5 |
+
- bn
|
| 6 |
+
metrics:
|
| 7 |
+
- wer
|
| 8 |
+
---
|
| 9 |
+
# thesven/whisper-tiny-bn-v2
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
## Model Description
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
This is the `openai/whisper-tiny` model fine-tuned on the OOD-Speech Bengali dataset. The fine-tuning was performed to enhance the model's performance on Bengali speech recognition tasks, especially for out-of-distribution samples.
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
## Model Details
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
|
| 24 |
+
- **Original Model:** [openai/whisper-tiny on Hugging Face](https://huggingface.co/openai/whisper-tiny)
|
| 25 |
+
|
| 26 |
+
- **Fine-tuned Model:** [thesven/whisper-tiny-bn-thesven on Hugging Face](https://huggingface.co/thesven/whisper-tiny-bn-thesven)
|
| 27 |
+
|
| 28 |
+
- **Model Architecture:** Whisper ASR
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
## Intended Use
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
This model is intended for automatic speech recognition (ASR) tasks in the Bengali language, especially for scenarios that involve out-of-distribution samples.
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
## Training Data
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
The model was fine-tuned using the OOD-Speech Bengali dataset:
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
- **Training Samples:** 10,000
|
| 48 |
+
|
| 49 |
+
- **Validation Samples:** 1,000
|
| 50 |
+
|
| 51 |
+
- **Dataset Source:** [OOD-Speech: A Large Bengali Speech Recognition Dataset for Out-of-Distribution Benchmarking](https://arxiv.org/abs/2305.09688)
|
| 52 |
+
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
## Fine-tuning Procedure
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
The model was fine-tuned using the OOD-Speech Bengali dataset with [specific hyperparameters, optimizer, learning rate, epochs, etc. if available].
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
## Licensing
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
This model is released under the MIT License. You are free to use, modify, distribute, or sell it under the terms of the license.
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
MIT License
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
Copyright (c) 2023 thesven
|
| 78 |
+
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
| 82 |
+
|
| 83 |
+
of this software and associated documentation files (the "Software"), to deal
|
| 84 |
+
|
| 85 |
+
in the Software without restriction, including without limitation the rights
|
| 86 |
+
|
| 87 |
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
| 88 |
+
|
| 89 |
+
copies of the Software, and to permit persons to whom the Software is
|
| 90 |
+
|
| 91 |
+
furnished to do so, subject to the following conditions:
|
| 92 |
+
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
The above copyright notice and this permission notice shall be included in all
|
| 96 |
+
|
| 97 |
+
copies or substantial portions of the Software.
|
| 98 |
+
|
| 99 |
+
|
| 100 |
+
|
| 101 |
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
| 102 |
+
|
| 103 |
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
| 104 |
+
|
| 105 |
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
| 106 |
+
|
| 107 |
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 108 |
+
|
| 109 |
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 110 |
+
|
| 111 |
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 112 |
+
|
| 113 |
+
SOFTWARE.
|
| 114 |
+
|
| 115 |
+
```
|
| 116 |
+
|
| 117 |
+
|
| 118 |
+
|
| 119 |
+
|
| 120 |
+
## Citation
|
| 121 |
+
|
| 122 |
+
|
| 123 |
+
|
| 124 |
+
```
|
| 125 |
+
|
| 126 |
+
@article{OOD-Speech2023,
|
| 127 |
+
|
| 128 |
+
title={OOD-Speech: A Large Bengali Speech Recognition Dataset for Out-of-Distribution Benchmarking},
|
| 129 |
+
|
| 130 |
+
author={Authors of the OOD-Speech paper},
|
| 131 |
+
|
| 132 |
+
journal={arXiv preprint arXiv:2305.09688},
|
| 133 |
+
|
| 134 |
+
year={2023}
|
| 135 |
+
|
| 136 |
+
}
|
| 137 |
+
|
| 138 |
+
```
|