Language model biasing system

1. A computer-implemented method comprising:

obtaining a dialog state of a user device associated with a user;

before the user speaks an utterance to the user device:

identifying, from an n-gram cache, an initial set of n-grams that represent one or more words or phrases corresponding to the dialog state of the user device; and

biasing a language model on the initial set of n-grams to increase a likelihood of output of the one or more words or phrases corresponding to the dialog state of the user device;

receiving audio data indicating the utterance of the user;

processing the audio data using the biased language model to generate a transcription of the utterance; and

providing the transcription of the utterance for output.

2. The method of claim 1 , further comprising:

receiving context data for the user device

wherein obtaining the dialog state comprises identifying the dialog state based on the context data.

3. The method of claim 2 , wherein the utterance is detected by the user device providing an interface to the user, and wherein the context data comprises data that indicates a topic corresponding to the interface.

4. The method of claim 2 , wherein the utterance is detected by the user device providing an interface to the user, and wherein the context data comprises data indicating a task to be performed using the interface.

5. The method of claim 2 , wherein the utterance is detected by the user device providing an interface to the user, and wherein the context data comprises data indicating a step for completing a portion of task to be performed using the interface.

6. The method of claim 2 , wherein the context data indicates one or more words or phrases included in a graphical user interface of the user device at a time that the utterance was spoken.

7. The method of claim 2 , wherein:

the context data comprises an application identifier or dialog state identifier; and

identifying the initial set of n-grams that represent the one or more words or phrases comprises retrieving data indicating one or more words or phrases corresponding to the application identifier or dialog state identifier.

8. The method of claim 2 , wherein identifying the initial set of n-grams that represent one or more words or phrases corresponding to the dialog state of the user device comprises:

identifying a first set of one or more n-grams from the context data; and

generating an expanded set of one or more n-grams based at least on the first set of n-grams, the expanded set of n-grams comprising one or more n-grams that are different from the n-grams in the first set of n-grams.

9. The method of claim 1 , wherein identifying the dialog state of the user device comprises identifying one of a plurality of different dialog states of the user device that each correspond to a different interface or view of an application.

10. The method of claim 9 , wherein each of the plurality of different dialog states of the user device is associated with a predetermined set of n-grams.

11. The method of claim 10 , wherein for at least one of the dialog states of the user device, one or more of the n-grams in the predetermined set of n-grams for the dialog state of the user device are not displayed by the application during the dialog state of the user device.

12. The method of claim 10 , wherein at least some of the n-grams in the predetermined set of n-grams are designated by a developer of the application.

13. The method of claim 1 , wherein receiving the audio data comprises receiving, by a server system, audio data provided by the user device over a communication network.

14. The method of claim 13 , wherein identifying the dialog state of the user device comprises identifying, by the server system, the dialog state of the user device based on additional data provided by the user device over the communication network.

15. The method of claim 14 , wherein the additional data comprises a dialog state identifier provided by the user device or an application identifier provided by the user device.

16. The method of claim 15 , wherein determining the one or more words or phrases corresponding to the dialog state of the user device comprises retrieving the one or more words or phrases from a data repository based on the dialog state identifier or the application identifier.

17. A system comprising:

one or more computers; and

one or more computer-readable media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:

obtaining a dialog state of a user device associated with a user;

before the user speaks an utterance to the user device:

identifying, from an n-gram cache, an initial set of n-grams that represent one or more words or phrases corresponding to the dialog state of the user device; and

biasing a language model on the initial set of n-grams to increase a likelihood of output of the one or more words or phrases corresponding to the dialog state of the user device;

receiving audio data indicating the utterance of the user;

processing the audio data using the biased language model to generate a transcription of the utterance; and

providing the transcription of the utterance for output.

18. One or more non-transitory computer-readable media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

obtaining a dialog state of a user device associated with a user;

before the user speaks an utterance to the user device:

identifying, from an n-gram cache, an initial set of n-grams that represent one or more words or phrases corresponding to the dialog state of the user device; and

biasing a language model on the initial set of n-grams to increase a likelihood of output of the one or more words or phrases corresponding to the dialog state of the user device;

receiving audio data indicating the utterance of the user;

processing the audio data using the biased language model to generate a transcription of the utterance; and

providing the transcription of the utterance for output.