Photo by

What is “confidence measures”?

evaluate reliability of recognition results.

Why measure confidence or reliability?

ambient noises, speaker variations, channel distortions and other mismatches.

Major methods in category?

  1. combination of predictor features;
    • log-likelihoods,
    • word duration
    • word graph density
  2. posterior probability: estimate the probability of the sequence given the acoustic observations.
  3. hypothesis testing: likelihood ratio = (word likelihood score) / (background model + competing model)
  4. utterance verification.

Specific Name

  • non-keyword rejection or keyword recognition: detect if the input speech does not contain any of the words in the recognizer vocabulary set.

  1. Out-Of-Vocabulary Detection and Confidence Measures for Speech Recognition Using Phone Models.
  2. Confidence Estimation for Machine Translation