TingChen-ppmc's picture
Update README.md
6e01064
metadata
language:
  - ch
license: apache-2.0
base_model: openai/whisper-small
datasets:
  - TingChen-ppmc/Changsha_Dialect_Conversational_Speech_Corpus
metrics:
  - wer
  - cer

This is a finetuned version of Whisper-small on Changsha_Dialect_Conversational_Speech_Corpus

Results

CER: 37.77

WER: 66.84

Chinese Word Error Rate (WER) is based on Jieba Word segmentation