Keyword detection modeling using contextual information
Features are disclosed for detecting words in audio using contextual information in addition to automatic speech recognition results. A detection model can be generated and used to determine whether a particular word, such as a keyword or “wake word,” has been uttered. The detection model can operate on features derived from an audio signal, contextual information associated with generation of the audio signal, and the like. In some embodiments, the detection model can be customized for particular users or groups of users based usage patterns associated with the users.
1. A system comprising:
computer-readable memory storing executable instructions; and
one or more processors in communication with the computer-readable memory, wherein the one or more processors are programmed by the executable instructions to at least:
receive audio data from a computing device subsequent to the computing device determining a wake word was detected in a first portion of the audio data;
generate speech recognition result data using at least the first portion of audio data and a second portion of the audio data;
generate acoustic data representing an acoustic property of a voice represented by the audio data;
generate contextual data representing a context associated with the audio data;
generate feature data using the speech recognition result data, acoustic data, and contextual data;
determine, using a statistical wake word detection model configured to receive the feature data as input, that the audio data fails to satisfy a detection criterion related to detecting a representation of the wake word; and
close an audio data stream from the computing device.
2. The system of claim 1 , wherein the one or more processors are further programmed by the executable instructions to determine the acoustic property, wherein the acoustic property comprises at least one of: prosody, energy, and speaking rate.
3. The system of claim 1 , wherein the contextual data represents at least one of: a time at which the audio data was generated, a geographic location of an audio input device, or a physical orientation of the audio input device.
4. The system of claim 1 , wherein the one or more processors are further programmed by the executable instructions to:
generate detection score data using the statistical wake word detection model and the feature data; and
determine that the detection score data fails to satisfy a detection threshold, wherein the detection score data failing to satisfy the detection threshold indicates the audio data fails to satisfy the detection criterion.
5. The system of claim 1 , wherein the one or more processors are further programmed by the executable instructions to receive identity data representing at least one of a start position or an end position of the first portion of audio data within the audio data.
6. The system of claim 5 , wherein the executable instructions to determine that the audio data fails to satisfy the detection criterion comprise executable instructions to determine that a portion of the audio data comprising at least one of the start position or the end position does not include a representation of the wake word.
7. The system of claim 1 , wherein the one or more processors are further programed to determine that the speech recognition result data represents a user intent.
8. A computer-implemented method comprising:
as implemented by a computing system comprising one or more processors configured to execute specific instructions,
receiving audio data representing audio of an utterance from a computing device subsequent to the computing device determining a wake word was detected in a first portion of the audio data;
generating speech recognition result data using at least the first portion of the audio data and a second portion of audio data;
generating acoustic data representing an acoustic property of a voice represented by the audio data;
generating contextual data representing a context associated with the audio data;
generating feature data using the speech recognition result data, acoustic data, and contextual data;
determining, using a statistical wake word detection model configured to receive the feature data as input, that the audio data fails to satisfy a detection criterion related to detecting a representation of the wake word; and
closing an audio data stream from the computing device.
9. The computer-implemented method of claim 8 , further comprising determining the acoustic property, wherein the acoustic property comprises at least one of: prosody, energy, and speaking rate.
10. The computer-implemented method of claim 8 , wherein determining that the audio data fails to satisfy the detection criterion comprises comparing a detection score to a detection threshold, wherein the audio data failing to satisfy the detection criterion comprises the detection score falling below the detection threshold.
11. The computer-implemented method of claim 8 , wherein generating the contextual data is based on at least one of: a time at which the audio data was generated, a geographic location of the computing device, and a physical orientation of the computing device.
12. A system comprising:
computer-readable memory storing executable instructions; and
one or more processors in communication with the computer-readable memory, wherein the one or more processors are programmed by the executable instructions to at least:
receive audio input comprising a plurality of portions of audio data from a computing device subsequent to the computing device determining a wake word was detected in a first portion of the plurality of portions of audio data, wherein fewer than all of the plurality of portions audio data include a representation of the wake word;
generate speech recognition result data using at least the first portion of audio data and a second portion of the plurality of portions of audio data;
generate acoustic data representing an acoustic property of a voice represented by the audio input;
generate contextual data representing a context associated with the audio input;
generate feature data using the speech recognition result data, acoustic data, and contextual data;
determine, using a statistical detection model configured to receive the feature data as input, that the first portion of audio data includes a representation of the wake word; and
perform an action corresponding to a request in the audio input.
13. The system of claim 12 , wherein the contextual data represents at least one of: a time at which the audio input was generated, a geographic location of the computing device, and a physical orientation of the computing device.
14. The system of claim 12 , wherein the one or more processors are further programmed by the executable instructions to analyze the contextual data against one or more rules of a logic network-based detection model.
15. The system of claim 12 , wherein the second portion of the audio data comprises a representation of a spoken command, and wherein the one or more processors are further programmed to determine that the speech recognition result data represents the spoken command.