lucio commited on
Commit
118e13d
1 Parent(s): 03aef7f

update LM eval results

Browse files
README.md CHANGED
@@ -20,10 +20,16 @@ model-index:
20
  type: mozilla-foundation/common_voice_8_0
21
  args: ky
22
  metrics:
23
- - name: Test WER
 
 
 
 
 
 
24
  type: wer
25
  value: 31.28
26
- - name: Test CER
27
  type: cer
28
  value: 7.66
29
  ---
@@ -45,6 +51,8 @@ For a description of the model architecture, see [facebook/wav2vec2-xls-r-300m](
45
 
46
  The model vocabulary consists of the cyrillic alphabet with punctuation removed.
47
 
 
 
48
  ## Intended uses & limitations
49
 
50
  This model is expected to be of some utility for low-fidelity use cases such as:
 
20
  type: mozilla-foundation/common_voice_8_0
21
  args: ky
22
  metrics:
23
+ - name: Test WER (with LM)
24
+ type: wer
25
+ value: 19.01
26
+ - name: Test CER (with LM)
27
+ type: cer
28
+ value: 5.38
29
+ - name: Test WER (no LM)
30
  type: wer
31
  value: 31.28
32
+ - name: Test CER (no LM)
33
  type: cer
34
  value: 7.66
35
  ---
 
51
 
52
  The model vocabulary consists of the cyrillic alphabet with punctuation removed.
53
 
54
+ The kenlm language model is built using the text of the train and invalidated corpus splits.
55
+
56
  ## Intended uses & limitations
57
 
58
  This model is expected to be of some utility for low-fidelity use cases such as:
mozilla-foundation_common_voice_8_0_ky_test_eval_results.txt CHANGED
@@ -1,2 +1,2 @@
1
- WER: 0.27494497432134996
2
- CER: 0.0674079866403361
 
1
+ WER: 0.19011371973587674
2
+ CER: 0.05388927913480272