TingChen-ppmc's picture
Update README.md
93e7ab0
metadata
language:
  - ch
license: apache-2.0
base_model: openai/whisper-small
datasets:
  - TingChen-ppmc/Tianjin_Dialect_Conversational_Speech_Corpus
metrics:
  - wer
  - cer

This is a finetuned version of Whisper-small on Tianjin_Dialect_Conversational_Speech_Corpus

Results

CER: 46.17

WER: 67.65

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