omarxadel commited on
Commit
a0db2b1
1 Parent(s): 3390a2d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +32 -36
README.md CHANGED
@@ -1,46 +1,42 @@
1
  ---
 
 
2
  tags:
3
- - generated_from_trainer
 
 
 
 
 
 
 
 
 
4
  model-index:
5
- - name: wav2vec2-large-xlsr-53-arabic-egyptian
6
- results: []
 
 
 
 
 
 
 
 
 
 
7
  ---
8
 
9
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
10
- should probably proofread and complete it, then remove this comment. -->
11
 
12
- # wav2vec2-large-xlsr-53-arabic-egyptian
13
 
14
- This model was trained from scratch on an unknown dataset.
15
 
16
- ## Model description
 
 
17
 
18
- More information needed
19
 
20
- ## Intended uses & limitations
21
-
22
- More information needed
23
-
24
- ## Training and evaluation data
25
-
26
- More information needed
27
-
28
- ## Training procedure
29
-
30
- ### Training hyperparameters
31
-
32
- The following hyperparameters were used during training:
33
- - learning_rate: 5e-05
34
- - train_batch_size: 8
35
- - eval_batch_size: 8
36
- - seed: 42
37
- - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
38
- - lr_scheduler_type: linear
39
- - num_epochs: 3.0
40
-
41
- ### Framework versions
42
-
43
- - Transformers 4.20.1
44
- - Pytorch 1.11.0+cu113
45
- - Datasets 2.3.2
46
- - Tokenizers 0.12.1
 
1
  ---
2
+ language: "ar"
3
+ pipeline_tag: automatic-speech-recognition
4
  tags:
5
+ - CTC
6
+ - Attention
7
+ - pytorch
8
+ - Transformer
9
+ license: "cc-by-nc-4.0"
10
+ datasets:
11
+ - MGB-3
12
+ - egyptian-arabic-conversational-speech-corpus
13
+ metrics:
14
+ - wer
15
  model-index:
16
+ - name: omarxadel/hubert-large-arabic-egyptian
17
+ results:
18
+ - task:
19
+ name: Automatic Speech Recognition
20
+ type: automatic-speech-recognition
21
+ metrics:
22
+ - name: Test WER
23
+ type: wer
24
+ value: 29.3755
25
+ - name: Validation WER
26
+ type: wer
27
+ value: 29.1828
28
  ---
29
 
30
+ # Wav2Vec2-XLSR-53 - with CTC fine-tuned on MGB-3 and Egyptian Arabic Conversational Speech Corpus (No LM)
 
31
 
32
+ This model is a fine-tuned version of [Wav2Vec2-XLSR-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53). We finetuned this model on the MGB-3 and Egyptian Arabic Conversational Speech Corpus datasets, acheiving WER of `29.3755%`.
33
 
34
+ The performance of the model on the datasets is the following:
35
 
36
+ | Valid WER | Test WER |
37
+ |:---------:|:--------:|
38
+ | 29.18 | 29.37 |
39
 
40
+ # Acknowledgement
41
 
42
+ Model fine-tuning and data processing for this work were performed as a part of a Graduation Project from Faculty of Engineering, Alexandria University, CCE Program.