Unsupervised and active learning in automatic speech recognition for call classification

A system comprising:

• a processor;
• an automatic speech recognizer configured to control the processor to automatically transcribe utterance data not having a corresponding manual transcription based on a bootstrap model and produce a set of automatically transcribed data, wherein the bootstrap model is based on text data mined from a website relevant to a specific domain;
• a learning module configured to control the processor to select a predetermined number of utterances from the set of automatically transcribed data based on a geometrically computed n-tuple confidence score to yield a selected number of utterances, the selected number of predetermined utterances to be manually transcribed, added to a set of manually transcribed data, and deleted from the set of automatically transcribed data;
• a training module configured to control the processor to train a language model using the set of manually transcribed data and the set of automatically transcribed data;
• and
• a labeler to label a portion of one of the set of automatically transcribed data and the set of manually transcribed data.

A system comprising:

• a processor;
• and
• a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:
• performing automatic speech recognition using a bootstrap model on utterances of a plurality of utterance data, the utterances not having a corresponding manual transcription, to produce automatically transcribed utterances, wherein the bootstrap data is based on text data mined from a website relevant to a specific domain;
• training a language model using manually transcribed data and the automatically transcribed utterances;
• selecting, based on a geometrically computed n-tuple confidence score, for manual transcription, a predetermined number of utterances not having a corresponding manual transcription from the utterance data, to yield new manually transcribed data;
• and
• labelling one of the automatically transcribed utterances and one of the new manually transcribed data.

A computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:

• performing automatic speech recognition using a bootstrap model on utterances of a plurality of utterance data not having a corresponding manual transcription to produce automatically transcribed utterances, wherein the bootstrap model is based on text data mined from a website relevant to a specific domain;
• training a model using manually transcribed data and the automatically transcribed utterances;
• selecting, based on a geometrically computed n-tuple confidence score, for manual transcription, a predetermined number of utterances, from the utterance data, not having a corresponding manual transcription to yield new manually transcribed data;
• receiving the new manually transcribed data;
• and
• permitting labeling of one of the automatically transcribed utterances and one of the new manually transcribed data and the manually transcribed data.

A method comprising:

• mining audio data from a source, to yield mined audio data;
• providing a bootstrap model, wherein the bootstrap model is based on text data of a website associated with a specific domain;
• and
• training a language model for call classification from the mined audio data based on the bootstrap model and a predetermined number of utterances not having a corresponding manual transcription data to yield a produced language model, the training being based on a geometrically computed n-tuple confidence score.

A computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:

• mining audio data from a source, to yield mined audio data;
• obtaining a bootstrap model, wherein the bootstrap model is based on text data of a website associated with a specific domain;
• and
• training a language model for call classification from the mined audio data based on the bootstrap model and a predetermined number of utterances not having a corresponding manual transcription data to yield a produced language model, the training being based at least in part on a geometrically computed n-tuple confidence score.

An apparatus comprising:

• a processor;
• and
• a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:
• mining audio data from a source, to yield mined audio data;
• obtaining a bootstrap model, wherein the bootstrap model is based on text data of a website associated with a specific domain;
• and
• training a language model for call classification from the audio data based on the bootstrap model and a predetermined number of utterances not having a corresponding manual transcription data, to yield a produced language model, the training based on a geometrically computed n-tuple confidence score to produce the language model.