eesungkim commited on
Commit
fdc8412
1 Parent(s): d3e710f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +4 -3
README.md CHANGED
@@ -76,7 +76,7 @@ Conformer-Transducer model is an autoregressive variant of Conformer model [2] f
76
 
77
  The model was finetuned based on the pre-trained English Model for over several epochs.
78
 
79
- There are several transcripting and sub-word modeling methods for Korean speech recognition. This model uses sentencepiece subwords of Hangul characters based on phonetic transcription using Google Sentencepiece Tokenizer [3].
80
 
81
  ### Datasets
82
 
@@ -90,8 +90,9 @@ v1.7.0rc | SentencePiece Char | 6.94% | 7.38% | 19.49% | 22.73%
90
 
91
  ## Limitations
92
 
93
- Eg:
94
- Since this model was trained on publically available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.
 
95
 
96
 
97
  ## References
 
76
 
77
  The model was finetuned based on the pre-trained English Model for over several epochs.
78
 
79
+ There are several transcribing and sub-word modeling methods for Korean speech recognition. This model uses sentencepiece subwords of Hangul characters based on phonetic transcription using Google Sentencepiece Tokenizer [3].
80
 
81
  ### Datasets
82
 
 
90
 
91
  ## Limitations
92
 
93
+ Since this model was trained on publically available speech datasets, the performance of this model might degrade for speech which including technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.
94
+
95
+ This model produces a spoken-form token sequence. If you want to have a written form, you can consider applying inverse text normalization.
96
 
97
 
98
  ## References