System and method for unsupervised and active learning for automatic speech recognition
View Patent ↗A system and method is provided for combining active and unsupervised learning for automatic speech recognition. This process enables a reduction in the amount of human supervision required for training acoustic and language models and an increase in the performance given the transcribed and un-transcribed data.
1. A computer-implemented method performed by a processor, the method comprising:
identifying, in a database of utterances, transcribed utterances and un-transcribed utterances;
ordering, via the processor, transcription candidate utterances from the un-transcribed utterances based on confidence scores of the transcription candidate utterances, to yield a selectively sampled order;
transcribing, via the processor, a top n utterances from the selectively sampled order, to yield additional transcribed utterances and remainder un-transcribed utterances, wherein the remainder un-transcribed utterances are the un-transcribed utterances without the additional transcribed utterances;
receiving human-transcribed utterances, wherein the human-transcribed utterances are selected from a bottom k utterances from the selectively sampled order for human transcription based on the confidence scores;
adding the additional transcribed utterances and the human-transcribed utterances to the database of utterances; and
training acoustic and language models using the additional transcribed utterances and the human-transcribed utterances.
2. The method of claim 1 , further comprising continuing the identifying, the transcribing, and the adding until a word error rate has converged.
3. The method of claim 2 , further comprising determining the confidence scores using an acoustic model.
4. The method of claim 2 , further comprising determining the confidence scores using a language model.
5. The method of claim 1 , further comprising:
upon adding the additional transcribed utterances to the database of utterances,
removing the additional utterances from the un-transcribed utterances.
6. The method of claim 1 , wherein the confidence scores of the transcription candidate utterances are associated with an arithmetic mean of confidences scores of words contained within each transcription candidate utterance.
7. The method of claim 1 , wherein the adding of the additional transcribed utterances and the human-transcribed utterances to the database of utterances is used to create an automatic speech recognition module.
8. A system comprising:
a processor; and
a computer-readable storage medium having instructions stored which, when executed by the processor, result in the processor performing operations comprising:
identifying, in a database of utterances, transcribed utterances and untranscribed utterances;
ordering, via the processor, transcription candidate utterances from the un-transcribed utterances based on confidence scores of the transcription candidate utterances, to yield a selectively sampled order;
transcribing, via the processor, a top n utterances from the selectively sampled order, to yield additional transcribed utterances and remainder un-transcribed utterances, wherein the remainder un-transcribed utterances are the un-transcribed utterances without the additional transcribed utterances;
receiving human-transcribed utterances, wherein the human-transcribed utterances are selected from a bottom k utterances from the selectively sampled order for human transcription based on the confidence scores;
adding the additional transcribed utterances and the human-transcribed utterances to the database of utterances; and
training acoustic and language models using the additional transcribed utterances and the human-transcribed utterances.
9. The system of claim 8 , the computer-readable storage medium having additional instructions stored which, when executed by the processor, result in operations comprising continuing the identifying, the transcribing, and the adding until a word error rate has converged.
10. The system of claim 9 , the computer-readable storage medium having additional instructions stored which, when executed by the processor, result in operations comprising determining the confidence scores using an acoustic model.
11. The system of claim 9 , the computer-readable storage medium having additional instructions stored which, when executed by the processor, result in operations comprising determining the confidence scores using a language model.
12. The system of claim 8 , the computer-readable storage medium having additional instructions stored which, when executed by the processor, result in operations comprising:
upon adding the additional transcribed utterances to the database of utterances,
removing the additional utterances from the un-transcribed utterances.
13. The system of claim 8 , wherein the confidence scores of the transcription candidate utterances are associated with an arithmetic mean of confidences scores of words contained within each transcription candidate utterance.
14. The system of claim 8 , wherein the adding of the additional transcribed utterances and the human-transcribed utterances to the database of utterances is used to create an automatic speech recognition module.
15. A computer-readable storage device having instructions stored which, when executed by a computing device, result in the computing device performing operations comprising:
identifying, in a database of utterances, transcribed utterances and un-transcribed utterances;
ordering, via the processor, transcription candidate utterances from the un-transcribed utterances based on confidence scores of the transcription candidate utterances, to yield a selectively sampled order;
transcribing, via the processor, a top n utterances from the selectively sampled order, to yield additional transcribed utterances and remainder un-transcribed utterances, wherein the remainder un-transcribed utterances are the un-transcribed utterances without the additional transcribed utterances;
receiving human-transcribed utterances, wherein the human-transcribed utterances are selected from a bottom k utterances from the selectively sampled order for human transcription based on the confidence scores;
adding the additional transcribed utterances and the human-transcribed utterances to the database of utterances; and
training acoustic and language models using the additional transcribed utterances and the human-transcribed utterances.
16. The computer-readable storage device of claim 15 , having additional instructions stored which, when executed by the processor, result in operations comprising continuing the identifying, the transcribing, and the adding until a word error rate has converged.
17. The computer-readable storage device of claim 16 , having additional instructions stored which, when executed by the processor, result in operations comprising determining the confidence scores using an acoustic model.
18. The computer-readable storage device of claim 16 , having additional instructions stored which, when executed by the processor, result in operations comprising determining the confidence scores using a language model.
19. The computer-readable storage device of claim 15 , having additional instructions stored which, when executed by the processor, result in operations comprising:
upon adding the additional transcribed utterances to the database of utterances,
removing the additional utterances from the un-transcribed utterances.
20. The computer-readable storage device of claim 15 , wherein the confidence scores of the transcription candidate utterances are associated with an arithmetic mean of confidences scores of words contained within each transcription candidate utterance.