metadata
language:
- ch
license: apache-2.0
base_model: openai/whisper-small
datasets:
- TingChen-ppmc/Zhengzhou_Dialect_Conversational_Speech_Corpus
metrics:
- wer
- cer
This is a finetuned version of Whisper-small on Zhengzhou_Dialect_Conversational_Speech_Corpus
Results
CER: 80.95
WER: 79.70
Chinese Word Error Rate (WER) is based on Jieba Word segmentation