Confidence; AM/LM scores

[Rafael Setyan]

Hi there,

Currently I am developing the ASR system with Kaldi. As I understand it, the LM-score how Dan mentioned here(https://groups.google.com/g/kaldi-help/c/IIRb5L_zRew) is the negative log probability, AM-score is log probability(of the acoustic sequences) and likelihood is the sum of this two. The AM/LM are calculated with  GetLinearSymbol method(https://github.com/kaldi-asr/kaldi/blob/master/src/online2bin/online2-wav-nnet3-latgen-faster.cc#L54) and likelihood is calculated in (https://github.com/kaldi-asr/kaldi/blob/master/src/online2bin/online2-wav-nnet3-latgen-faster.cc#L56).

I would like to know is there any recipe/formula for getting the confidence from AM/LM score for given utterance, that can be used as a decision boundary?

I have found some 3rd party code of calculating the confidence, however there were no proper documentation. (https://github.com/skit-ai/kaldi-serve/blob/master/include/kaldiserve/decoder.hpp#L194)

Thanks in advance