Confidence Estimation and Deletion Prediction Using Bidirectional Recurrent Neural Networks
This work addresses deletion errors in speech recognition for limited-resource and mismatched conditions, which is an incremental improvement over standard confidence estimation schemes.
The paper tackled the problem of deletion errors in automatic speech recognition by using bidirectional recurrent neural networks to estimate confidence scores for both predicted and deleted words, resulting in improved data selection that favored transcriptions with lower deletion errors in experiments on IARPA Babel/Material program languages.
The standard approach to assess reliability of automatic speech transcriptions is through the use of confidence scores. If accurate, these scores provide a flexible mechanism to flag transcription errors for upstream and downstream applications. One challenging type of errors that recognisers make are deletions. These errors are not accounted for by the standard confidence estimation schemes and are hard to rectify in the upstream and downstream processing. High deletion rates are prominent in limited resource and highly mismatched training/testing conditions studied under IARPA Babel and Material programs. This paper looks at the use of bidirectional recurrent neural networks to yield confidence estimates in predicted as well as deleted words. Several simple schemes are examined for combination. To assess usefulness of this approach, the combined confidence score is examined for untranscribed data selection that favours transcriptions with lower deletion errors. Experiments are conducted using IARPA Babel/Material program languages.