NATURAL MACHINE CONVERSING METHOD AND APPARATUS

• US 20190035391A1
• Filed: 07/27/2017
• Published: 01/31/2019
• Est. Priority Date: 07/27/2017
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Danielescu, Lavinia A., Nikkila, Shawn C., Firby, Robert J., Hockey, Beth Ann
• Granted Patent: US 10,360,909 B2
• Time in Patent Office: Days
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

Apparatuses, methods and storage medium associated with a spoken dialogue system are disclosed herein. In embodiments, an apparatus for natural machine conversing with a user may comprise a listening component to detect a keyword that denotes start of a conversation; a dialogue engine to converse with the user during the conversation; and a controller to selectively activate or cause to be activated one of the listening component or the dialogue component, and to pass control to the activated listening component or the activated dialogue engine, based at least in part on a state of the conversation. Other embodiments may be disclosed or claimed.

Claims

1. An apparatus for natural machine conversing with a user, comprising:

listening circuitry, including a first autonomous speech recognition (ASR) engine, to detect a keyword that denotes start of a conversation between the apparatus and the user;

dialogue circuitry, including a second ASR engine with greater speech recognition capability than the first ASR engine, to converse with the user during the conversation, including to detect for one or more conversation continuing features during the conversation, the one or more conversation continuing features include at least selected pitch attributes of the user; and

a controller coupled to the listening circuitry and the dialogue circuitry to selectively activate or cause to be activated one of the listening circuitry or the dialogue circuitry, and to pass control to the activated listening circuitry or the activated dialogue circuitry, based at least in part on a current state of the conversation, which includes at least a start state, a continuing state, and an end state;

wherein the dialogue circuitry is to further notify the controller of the continuing state, on detection of at least one conversation continuing features to keep the dialogue engine from being prematurely deactivated prior to end of the conversation with the user;

wherein detection for conversation continuing features include detection for at least selected pitch attributes of the user.

2. The apparatus of claim 1, wherein the listening circuitry during operation consumes less power than power consumed by the dialogue circuitry during its operation.

3. The apparatus of claim 1, wherein the listening circuitry is to send the controller a conversation state indicator that denotes a detection of the keyword or the start of the conversation, or the dialogue circuitry is to send the controller a conversation state indicator that denotes a desire to start the conversation.

4. The apparatus of claim 3, wherein the controller is to activate or cause to be activated the dialogue circuitry, and to pass control to the dialogue circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversation, or the conversation state indicator that denotes a desire of the dialogue circuitry to start the conversation.

5. The apparatus of claim 4, wherein the controller is to further deactivate or cause to be deactivated the listening circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversation, or on receipt of the conversation state indicator that denotes a desire to start the conversation.

6. The apparatus of claim 4, wherein the dialogue circuitry, on activation, is to set a timer, to recognize user utterance, and to respond to the user utterance.

7. The apparatus of claim 6, wherein the controller is to reset or cause the timer to be reset on receipt of a conversation state indicator that indicates continuation of the conversation.

8. The apparatus of claim 7, wherein the dialogue circuitry is to send, and the controller is to receive from the dialogue circuitry, a conversation state indicator that denotes the dialogue circuitry recognized user utterance, the dialogue circuitry responded to user utterance, or the dialogue circuitry detected one or more of the continuing conversation features.

9. The apparatus of claim 6, wherein the controller is to activate or cause to be activated the listening circuitry, and to pass control to the listening circuitry, on receipt of a conversation state indicator that denotes the state of the conversation is at the end state or an expiration of the timer.

10. The apparatus of claim 5, wherein the dialogue circuitry is to further detect for end of conversation features, and wherein the controller is to activate the listening circuitry, and to pass control to the listening circuitry, on receipt from the dialogue circuitry a conversation state indicator that denotes a detection of end of conversation features.

11. The apparatus of claim 1, wherein the controller is to activate the listening circuitry, and to pass control to the listening circuitry, on receipt of a conversation state indicator that denotes a detection of an end of the conversation.

12. A machine implemented method for natural conversing with a user, comprising:

setting, by dialogue circuitry of a computer device, a timer of the dialogue circuitry, on activation of the dialogue circuitry, the dialogue circuitry further including a first automatic speech recognition (ASR) engine;

on setting the timer, conversing, by the dialogue circuitry, with a user of the computer device; and

on expiration of the timer, notifying a controller of the computer device, by the dialogue circuitry, of the expiration of the timer;

wherein on notification of the expiration of the timer, the dialogue circuitry is de-activated, and a listening circuitry of the computer device is activated instead, the listening circuitry having a second ASR engine with lesser speech recognition capability than the first ASR;

wherein the method further comprises;

detecting, by the dialogue circuitry, for one or more conversation continuing features during the conversing, the one or more conversation continuing features include at least selected pitch attributes of the user;

on detection of at least one conversation continuing features, notifying the controller, by the dialogue circuitry of the detection; and

on notification of the detection, resetting the timer to keep the dialogue circuitry from prematurely deactivated prior to end of conversing with the user.

13. The method of claim 12, further comprising sending to the controller, by a listening circuitry of the computer device, a conversation state indicator that denotes a detection of a keyword or a start of conversing with the user, or sending to the controller, by the dialogue circuitry, a conversation state indicator that denotes a desire to start conversing with the user.

14. The method of claim 13, further comprising activating or causing to be activated, by the controller, the dialogue circuitry, and passing control to the dialogue circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversing with the user, or the conversation state indicator that denotes a desire of the dialogue circuitry to start conversing with the user.

15. The method of claim 14, further comprising deactivating or causing to be deactivated, by the controller, the listening circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversing with the user, or on receipt of the conversation state indicator that denotes a desire to start conversing with the user.

16. The method of claim 15, further comprising activating or causing to be activated, by the controller, the listening circuitry, and passing control to the listening circuitry, on receipt of a conversation state indicator that denotes the state of the conversation is at an end state or an expiration of the timer.

17. The methods of claim 15, further comprising detecting, by the dialogue circuitry, for end of conversation features, and activating, by the controller, the listening circuitry, and passing control to the listening circuitry, on receipt from the dialogue circuitry a conversation state indicator that denotes a detection of end of conversation features.

18. The method of claim 12, further comprising activating, by the controller, the listening circuitry, and passing control to the listening circuitry, on receipt of a conversation state indicator that denotes a detection of an end of the conversation.

19. One or more computer-readable medium (CRM) having instructions to cause a computer device, in response to the execution of the instructions, to:

analyze, with a controller of the computer device, indicators that denote states of a conversation between the computer device and a user, the states including a start conversation state, a conversation continuing state, and an end conversation state; and

selectively activate, causing to be activated, with the controller, a selected one of a listening circuitry or a dialogue circuitry of the computer device, based at least in part on the state of the conversation with the user, and pass control to the activated listening circuitry to detect for a keyword that denotes start of a conversation, or a dialogue circuitry to converse with the user, the listening circuitry including a first autonomous speech recognition (ASR) engine, and the dialogue circuitry including a second ASR with greater speech recognition capability than the first ASR;

wherein the dialogue circuitry, while conversing with the user, further detects for continuing conversation features that includes at least selected pitch attributes of the user, and on detection of one or more continuing conversation features, notifies the controller of the detection to prevent the dialogue circuitry from prematurely deactivated prior to an end of the conversation.

20. The CRM of claim 19, wherein the computer device is further caused to send, with the listening circuitry, to the controller a conversation state indicator that denotes a detection of the keyword or the start of the conversation, or the dialogue circuitry is to send the controller a conversation state indicator that denotes a desire to start the conversation.

21. The CRM of claim 20, wherein the computer device is further caused, with the controller, to activate the dialogue circuitry, and to pass control to the dialogue circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversation, or the conversation state indicator that denotes a desire of the dialogue circuitry to start the conversation.

22. The CRM of claim 21, wherein the computer device is further caused, with the controller to deactivate the listening circuitry, on receipt of the conversation state indicator that denotes a detection of the keyword or the start of the conversation, or on receipt of the conversation state indicator that denotes a desire to start the conversation.

23. The CRM of claim 21, wherein the computer device is further caused, with the dialogue circuitry, on activation, to set a timer, to recognize user utterance, and to respond to the user utterance.

24. The CRM of claim 23, wherein the computer device is further caused, with the controller, to reset the timer, on receipt of a conversation state indicator that indicates continuation of the conversation.

25. The CRM of claim 24, wherein the computer device is further caused, with the dialogue circuitry, to send to the controller, a conversation state indicator that denotes the dialogue circuitry recognized user utterance, the dialogue circuitry responded to user utterance, or the dialogue circuitry detected continuing conversation features.