Replies: 1 comment
-
|
Hi @Zhao-124! I'm Dosu and I’m helping the sherpa-onnx team. Sherpa-onnx provides several Chinese ASR models (like paraformer, zipformer, conformer), but their standard output does not include a confidence level or score—only recognized text, tokens, and sometimes timestamps are shown in the documentation and usage examples. There is no explicit mention of confidence scores being available for Chinese models as part of the recognition result in the official docs or output examples see here and here. Some features like hotword boosting and keyword spotting use scores internally, but these are specific to those tasks and not general ASR confidence scores see here. If you need confidence levels, you may need to modify the code to extract scores from the decoder or beam search outputs, as this is not currently exposed by default. To reply, just mention @dosu. How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I want to obtain the confidence level. Are there any Chinese models that can output the confidence level?
Beta Was this translation helpful? Give feedback.
All reactions