Language model biasing modulation
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for modulating language model biasing. In some implementations, context data is received. A likely context associated with a user is determined based on at least a portion of the context data. One or more language model biasing parameters based at least on the likely context associated with the user is selected. A context confidence score associated with the likely context based on at least a portion of the context data is determined. One or more language model biasing parameters based at least on the context confidence score is adjusted. A baseline language model based at least on the one or more of the adjusted language model biasing parameters is biased. The baseline language model is provided for use by an automated speech recognizer (ASR).
1. A computer-implemented method comprising:
receiving, by an automated speech recognizer (ASR) system that includes (i) a context selector, (ii) a language model biaser, (iii) an ASR, (iv) a particular language model, and (v) a previously biased language model, audio data corresponding to an utterance of a user;
determining, by the context selector of the ASR system, that the user is likely no longer within a particular context that is associated the previously biased language model;
in response to determining that the user is likely no longer within a particular context that is associated with the previously biased language model, selecting, by the language model biaser of the ASR system, the particular language model for use in transcribing utterances;
after selecting the baseline language model, generating, by the ASR of the ASR system, a transcription of the utterance using the particular language model; and
providing a representation of the transcription for output.
2. The method of claim 1 , wherein selecting the particular language model comprises returning the previously biased language model to a state, or reducing an effect of the previous biasing of the language model.
3. The method of claim 1 , wherein determining that the user is likely no longer within the particular context comprises determining that a predetermined period of time has elapsed.
4. The method of claim 1 , wherein determining that the user is likely no longer within the particular context comprises determining that the user is likely in a different context.
5. The method of claim 1 , wherein determining that the user is likely no longer within the particular context comprises determining that a confidence score associated with the user and the particular context does not satisfy a threshold.
6. The method of claim 1 , wherein determining that the user is likely no longer within the particular context comprises determining that the user is likely associated with another context that is inconsistent with the particular context.
7. The method of claim 1 , wherein selecting the particular language model comprises generating a modulated set of biasing parameters.
8. The method of claim 1 , wherein providing a representation of the transcription for output comprising providing a textual representation of the transcription for display on a graphical user interface.
9. A non-transitory computer storage medium storing instructions that, when executed by data processing apparatus, cause the one or more computers to perform operations comprising:
receiving, by an automated speech recognizer (ASR) system that includes (i) a context selector, (ii) a language model biaser, (iii) an ASR, (iv) a particular language model, and (v) a previously biased language model, audio data corresponding to an utterance of a user;
determining, by the context selector of the ASR system, that the user is likely no longer within a particular context that is associated the previously biased language model;
in response to determining that the user is likely no longer within a particular context that is associated with the previously biased language model, selecting, by the language model biaser of the ASR system, the particular language model for use in transcribing utterances;
after selecting the baseline language model, generating, by the ASR of the ASR system, a transcription of the utterance using the particular language model; and
providing a representation of the transcription for output.
10. The medium of claim 9 , wherein selecting the particular language model comprises returning the previously biased language model to a state, or reducing an effect of the previous biasing of the language model.
11. The medium of claim 9 , wherein determining that the user is likely no longer within the particular context comprises determining that a predetermined period of time has elapsed.
12. The medium of claim 9 , wherein determining that the user is likely no longer within the particular context comprises determining that the user is likely in a different context.
13. The medium of claim 9 , wherein determining that the user is likely no longer within the particular context comprises determining that a confidence score associated with the user and the particular context does not satisfy a threshold.
14. The medium of claim 9 , wherein determining that the user is likely no longer within the particular context comprises determining that the user is likely associated with another context that is inconsistent with the particular context.
15. The medium of claim 9 , wherein selecting the particular language model comprises generating a modulated set of biasing parameters.
16. An automated speech recognizer (ASR) system comprising:
one or more computers comprising (i) a context selector, (ii) a language model biaser, (iii) an ASR, (iv) a particular language model, and (v) a previously biased language model, and including one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations comprising:
receiving audio data corresponding to an utterance of a user;
determining, by the context selector, that the user is likely no longer within a particular context that is associated the previously biased language model;
in response to determining that the user is likely no longer within a particular context that is associated with the previously biased language model, selecting, by the language model biaser, the particular language model for use in transcribing utterances;
after selecting the baseline language model, generating, by the ASR, a transcription of the utterance using the particular language model; and
providing a representation of the transcription for output.
17. The system of claim 16 , wherein selecting the particular language model comprises returning the previously biased language model to a state, or reducing an effect of the previous biasing of the language model.
18. The system of claim 16 , wherein determining that the user is likely no longer within the particular context comprises determining that a predetermined period of time has elapsed.
19. The system of claim 16 , wherein determining that the user is likely no longer within the particular context comprises determining that the user is likely in a different context.
20. The system of claim 16 , wherein determining that the user is likely no longer within the particular context comprises determining that a confidence score associated with the user and the particular context does not satisfy a threshold.