In Deepspeech documentation, definition of confidence is:
Confidence is roughly the sum of the acoustic model logit values for each timestep/character that contributed to the creation of this transcription.
But on running on different audios, confidence comes like -5, -20. I want to know the confidence of the transcribed text, i.e., in percentage what are the chances that sentence is correct. Can the confidence by Deepspeech model be used to calculate the accuracy?