Automated speech recognition system for natural language understanding

• US 8,560,321 B1
• Filed: 01/05/2011
• Issued: 10/15/2013
• Est. Priority Date: 01/05/2011
• Status: Active Grant

• Current Assignee: Interactions, LLC
• Original Assignee: Interactions Corporation (Interactions, LLC)
• Inventors: Yeracaris, Yoryos, Gray, Philip M., Dreher, James P.
• Primary Examiner(s): McFadden, Susan
• Application Number:
  US12/985,174
• Time in Patent Office: 1,014 Days
• Field of Search: 704/270
• US Class Current: 704/270.1
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• 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 2015/0638 Interactive procedures
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00
• CPC Class Codes: H04M 3/5166 in combination with interactive voice response systems or voice portals, e.g. as front-ends

Abstract

An interactive response system directs input to a software-based router, which is able to intelligently respond to the input by drawing on a combination of human agents, advanced speech recognition and expert systems. The system utilizes human “intent analysts” for purposes of interpreting customer utterance. Automated speech recognition subsystems are trained by coupling utterances with IA-selected meanings for those utterances, using non-real time systems to develop the training information for the automated speech recognition subsystems.

Claims

1. A system for processing an interaction with a person, comprising:

a routing processor configured to receive data representing an utterance from the person;

an automated recognition processor in communication with the routing processor and configured to receive therefrom the data, the automated recognition processor producing responsive to the data an output substantially in real time corresponding to a determination of an intent of the person;

an analyst user interface device in communication with the routing processor, configured to present to a human analyst the utterance in perceptible form and to accept intent input from the human analyst; and

a training subsystem in communication with the routing processor, the automated recognition processor, and the intent input and producing in response training information for the automated recognition processor.

2. The system of claim 1, wherein the training subsystem comprises a second automated recognition processor not operating in real-time.

3. The system of claim 1, wherein the routing processor is configured to selectively use the automated recognition processor responsive to a quality of operation of the automated recognition processor.

4. The system of claim 1, further comprising a second analyst user interface device in communication with the routing processor, configured to present to a second human analyst the utterance in perceptible form and to accept second intent input from the second human analyst, wherein the intent input is selectively processed by the training subsystem responsive to the intent input matching the second intent input.

5. The system of claim 1, wherein the training subsystem comprises a plurality of non real-time ASRs, the training information being generated selectively upon concurrence of the non real-time ASRs.

6. The system of claim 1, wherein the training information comprises a statistical model corresponding to the utterance.

7. A computer-implemented method for operating an interactive response system comprising:

receiving data representing an utterance from a person;

using a processor, automatically presenting the utterance in perceptible form to an intent analyst through an analyst user interface;

accepting intent input from intent analyst through the analyst user interface, where the intent input characterizes the intent analyst'"'"'s interpretation of the person'"'"'s intent expressed in the utterance;

providing the data and the intent input to a training subsystem;

accepting from the training subsystem training information generated responsive to the data and the intent input; and

sending the training information to a real-time recognition processor in order to improve performance thereof.

8. The computer-implemented method of claim 7, further comprising accepting intent input from a second intent analyst, wherein said providing is done responsive to intent input from the intent analyst matching intent input from the second intent analyst.

9. The computer-implemented method of claim 7, wherein the training information is not generated in real-time by the training subsystem.

10. The computer-implemented method of claim 7, further comprising sending a second utterance to the real-time recognition processor responsive to performance of the real-time recognition processor exceeding a threshold quality level.

11. The computer-implemented method of claim 7, further comprising automatically presenting the utterance in perceptible form to a second analyst and accepting second intent input from the second analyst, wherein the intent input is selectively provided to the training subsystem responsive to the intent input matching the second intent input.

12. The computer-implemented method of claim 7, wherein the training subsystem comprises a plurality of non real-time ASRs, the training information being generated selectively upon concurrence of the non real-time ASRs.

13. The computer-implemented method of claim 7, wherein the training information comprises a statistical model corresponding to the utterance.

14. A system for managing interactions with a person, comprising non-transitory computer storage media storing programming instructions executable by at least one processor to:

receive data representing an utterance from a person;

automatically present the utterance in perceptible form to an intent analyst through an analyst user interface;

accept intent input from intent analyst through the analyst user interface, where the intent input characterizes the intent analyst'"'"'s interpretation of the person'"'"'s intent expressed in the utterance;

provide the data and the intent input to a training subsystem;

accept from the training subsystem training information generated responsive to the data and the intent input; and

send the training information to a real-time recognition processor in order to improve performance thereof.

15. The system of claim 14, wherein the instructions further comprise instructions to accept intent input from a second intent analyst, wherein said instructions to provide are responsive to intent input from the intent analyst matching intent input from the second intent analyst.

16. The system of claim 14, wherein the training information is not generated in real-time by the training subsystem.

17. The system of claim 14, wherein the instructions further comprise instructions to send a second utterance to the real-time recognition processor responsive to performance of the real-time recognition processor exceeding a threshold quality level.

18. The system of claim 14, wherein the instructions further comprise instructions to automatically present the utterance in perceptible form to a second analyst and accept second intent input from the second analyst, wherein the intent input is selectively provided to the training subsystem responsive to the intent input matching the second intent input.

19. The system of claim 14, wherein the training subsystem comprises a plurality of non real-time ASRs, the training information being generated selectively upon concurrence of the non real-time ASRs.

20. The system of claim 14, wherein the training information comprises a statistical model corresponding to the utterance.