Update README.md
Browse files
README.md
CHANGED
|
@@ -2,6 +2,8 @@
|
|
| 2 |
language: ru
|
| 3 |
datasets:
|
| 4 |
- SberDevices/Golos
|
|
|
|
|
|
|
| 5 |
metrics:
|
| 6 |
- wer
|
| 7 |
- cer
|
|
@@ -12,7 +14,7 @@ tags:
|
|
| 12 |
- xlsr-fine-tuning-week
|
| 13 |
license: apache-2.0
|
| 14 |
widget:
|
| 15 |
-
- example_title: test sound with Russian speech
|
| 16 |
src: https://huggingface.co/bond005/wav2vec2-large-ru-golos/resolve/main/test_sound_ru.flac
|
| 17 |
model-index:
|
| 18 |
- name: XLSR Wav2Vec2 Russian by Ivan Bondarenko
|
|
@@ -27,10 +29,10 @@ model-index:
|
|
| 27 |
metrics:
|
| 28 |
- name: Test WER
|
| 29 |
type: wer
|
| 30 |
-
value:
|
| 31 |
- name: Test CER
|
| 32 |
type: cer
|
| 33 |
-
value:
|
| 34 |
- task:
|
| 35 |
name: Speech Recognition
|
| 36 |
type: automatic-speech-recognition
|
|
@@ -41,10 +43,10 @@ model-index:
|
|
| 41 |
metrics:
|
| 42 |
- name: Test WER
|
| 43 |
type: wer
|
| 44 |
-
value:
|
| 45 |
- name: Test CER
|
| 46 |
type: cer
|
| 47 |
-
value:
|
| 48 |
- task:
|
| 49 |
name: Automatic Speech Recognition
|
| 50 |
type: automatic-speech-recognition
|
|
@@ -55,10 +57,10 @@ model-index:
|
|
| 55 |
metrics:
|
| 56 |
- name: Test WER
|
| 57 |
type: wer
|
| 58 |
-
value:
|
| 59 |
- name: Test CER
|
| 60 |
type: cer
|
| 61 |
-
value: 4.
|
| 62 |
- task:
|
| 63 |
name: Automatic Speech Recognition
|
| 64 |
type: automatic-speech-recognition
|
|
@@ -69,10 +71,10 @@ model-index:
|
|
| 69 |
metrics:
|
| 70 |
- name: Test WER
|
| 71 |
type: wer
|
| 72 |
-
value:
|
| 73 |
- name: Test CER
|
| 74 |
type: cer
|
| 75 |
-
value:
|
| 76 |
- task:
|
| 77 |
name: Automatic Speech Recognition
|
| 78 |
type: automatic-speech-recognition
|
|
@@ -83,10 +85,24 @@ model-index:
|
|
| 83 |
metrics:
|
| 84 |
- name: Test WER
|
| 85 |
type: wer
|
| 86 |
-
value:
|
| 87 |
- name: Test CER
|
| 88 |
type: cer
|
| 89 |
-
value:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
---
|
| 91 |
|
| 92 |
# Wav2Vec2-Large-Ru-Golos
|
|
@@ -188,13 +204,13 @@ print("Character error rate on the Farfield domain:", farfield_cer)
|
|
| 188 |
|
| 189 |
| "crowd" | "farfield" |
|
| 190 |
|---------|------------|
|
| 191 |
-
|
|
| 192 |
|
| 193 |
*Result (CER, %)*:
|
| 194 |
|
| 195 |
| "crowd" | "farfield" |
|
| 196 |
|---------|------------|
|
| 197 |
-
|
|
| 198 |
|
| 199 |
You can see the evaluation script on other datasets, including Russian Librispeech and SOVA RuDevices, on my Kaggle web-page https://www.kaggle.com/code/bond005/wav2vec2-ru-eval
|
| 200 |
|
|
|
|
| 2 |
language: ru
|
| 3 |
datasets:
|
| 4 |
- SberDevices/Golos
|
| 5 |
+
- bond005/sova_rudevices
|
| 6 |
+
- bond005/rulibrispeech
|
| 7 |
metrics:
|
| 8 |
- wer
|
| 9 |
- cer
|
|
|
|
| 14 |
- xlsr-fine-tuning-week
|
| 15 |
license: apache-2.0
|
| 16 |
widget:
|
| 17 |
+
- example_title: test sound with Russian speech "нейросети это хорошо"
|
| 18 |
src: https://huggingface.co/bond005/wav2vec2-large-ru-golos/resolve/main/test_sound_ru.flac
|
| 19 |
model-index:
|
| 20 |
- name: XLSR Wav2Vec2 Russian by Ivan Bondarenko
|
|
|
|
| 29 |
metrics:
|
| 30 |
- name: Test WER
|
| 31 |
type: wer
|
| 32 |
+
value: 10.144
|
| 33 |
- name: Test CER
|
| 34 |
type: cer
|
| 35 |
+
value: 2.168
|
| 36 |
- task:
|
| 37 |
name: Speech Recognition
|
| 38 |
type: automatic-speech-recognition
|
|
|
|
| 43 |
metrics:
|
| 44 |
- name: Test WER
|
| 45 |
type: wer
|
| 46 |
+
value: 20.353
|
| 47 |
- name: Test CER
|
| 48 |
type: cer
|
| 49 |
+
value: 6.030
|
| 50 |
- task:
|
| 51 |
name: Automatic Speech Recognition
|
| 52 |
type: automatic-speech-recognition
|
|
|
|
| 57 |
metrics:
|
| 58 |
- name: Test WER
|
| 59 |
type: wer
|
| 60 |
+
value: 18.548
|
| 61 |
- name: Test CER
|
| 62 |
type: cer
|
| 63 |
+
value: 4.000
|
| 64 |
- task:
|
| 65 |
name: Automatic Speech Recognition
|
| 66 |
type: automatic-speech-recognition
|
|
|
|
| 71 |
metrics:
|
| 72 |
- name: Test WER
|
| 73 |
type: wer
|
| 74 |
+
value: 25.410
|
| 75 |
- name: Test CER
|
| 76 |
type: cer
|
| 77 |
+
value: 7.965
|
| 78 |
- task:
|
| 79 |
name: Automatic Speech Recognition
|
| 80 |
type: automatic-speech-recognition
|
|
|
|
| 85 |
metrics:
|
| 86 |
- name: Test WER
|
| 87 |
type: wer
|
| 88 |
+
value: 21.872
|
| 89 |
- name: Test CER
|
| 90 |
type: cer
|
| 91 |
+
value: 4.469
|
| 92 |
+
- task:
|
| 93 |
+
name: Automatic Speech Recognition
|
| 94 |
+
type: automatic-speech-recognition
|
| 95 |
+
dataset:
|
| 96 |
+
name: Voxforge Ru
|
| 97 |
+
type: dangrebenkin/voxforge-ru-dataset
|
| 98 |
+
args: ru
|
| 99 |
+
metrics:
|
| 100 |
+
- name: Test WER
|
| 101 |
+
type: wer
|
| 102 |
+
value: 27.084
|
| 103 |
+
- name: Test CER
|
| 104 |
+
type: cer
|
| 105 |
+
value: 6.986
|
| 106 |
---
|
| 107 |
|
| 108 |
# Wav2Vec2-Large-Ru-Golos
|
|
|
|
| 204 |
|
| 205 |
| "crowd" | "farfield" |
|
| 206 |
|---------|------------|
|
| 207 |
+
| 10.144 | 20.353 |
|
| 208 |
|
| 209 |
*Result (CER, %)*:
|
| 210 |
|
| 211 |
| "crowd" | "farfield" |
|
| 212 |
|---------|------------|
|
| 213 |
+
| 2.168 | 6.030 |
|
| 214 |
|
| 215 |
You can see the evaluation script on other datasets, including Russian Librispeech and SOVA RuDevices, on my Kaggle web-page https://www.kaggle.com/code/bond005/wav2vec2-ru-eval
|
| 216 |
|