thesven commited on
Commit
822d589
1 Parent(s): 99f1b20

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +138 -0
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
+ ```