Method and system for voice recognition menu navigation with error prevention and recovery

• US 7,246,062 B2
• Filed: 04/08/2002
• Issued: 07/17/2007
• Est. Priority Date: 04/08/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: SBC Technology Resources, Inc. (AT&T, Inc.)
• Inventors: Bushey, Robert Randal, Smart, Tracy Leigh, Knott, Benjamin Anthony, Martin, John Mills
• Primary Examiner(s): ABEBE, DANIEL DEMELASH
• Application Number:
  US10/118,478
• Publication Number:
  US 20030191648A1
• Time in Patent Office: 1,926 Days
• Field of Search: 704/275, 704/246, 345/728
• US Class Current: 704/246
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A method and system for error prevention and recovery of voice activated navigation through a menu having plural nodes provides situation dependent utterance verification by relating confirmation to utterance determination confidence levels. In one embodiment, a high confidence level results in implicit confirmation, a medium confidence level results in explicit confirmation and a low confidence level results in a concise interrogative prompt of a single word that requests the user to repeat the utterance. In situations where voice recognition is difficult, dual modality with DTMF navigation is provided as an option for menu selections.

Claims

1. A method for navigating a voice recognition menu having plural nodes, the method comprising:

receiving a user utterance at a voice recognition service;

determining a requested node associated with the user utterance;

classifying the user utterance as having a high, medium or low confidence that the determined requested node matches the node requested by the user utterance;

advancing a high confidence utterance to a menu node associated with the high confidence utterance;

requesting confirmation of a medium confidence utterance by prompting the user with the determined requested node;

requesting the user to restate a low confidence utterances;

counting the number of attempts to determine the received user utterance; and

automatically directing the user to a DTMF menu after a predetermined number of attempts.

2. The method of claim 1 further comprising:

obtaining implicit confirmation of high confidence utterances before advancing to associated menu nodes.

3. The method of claim 2 wherein obtaining implicit confirmation further comprises:

stating the determined requested node;

advancing to the determined requested node if the user either fails to respond; and

requesting the user to restate the user utterance if the user refutes the determined requested node.

4. The method of claim 1 wherein requesting confirmation of a medium confidence utterance further comprises:

obtaining explicit confirmation of medium confidence utterances.

5. The method of claim 4 wherein obtaining explicit confirmation further comprises:

stating the determined requested node as query to the user;

advancing to the determined requested node if the user responds with affirmation; and

requesting the user to restate the user utterance if the user refutes the determined requested node.

6. The method of claim 1 wherein requesting the user to restate a low confidence utterance further comprises:

stating a precise interrogative prompt; and

re-receiving the user utterance from the user.

7. The method of claim 6 wherein the precise interrogative prompt comprises a single word stated as a query.

8. The method of claim 7 wherein stating the precise interrogative prompt comprises stating the word “

Pardon”

as a query.

9. The method of claim 1, wherein:

the received user utterance is associated with a particular one of the plural nodes; and

different ones of the plural nodes have different quantitative confidence level ranges such that the classification of the user utterance as a high, medium or low confidence utterance is based at least in part on the quantitative confidence level ranges corresponding with the particular node.

10. The method of claim 1, wherein each node of the voice recognition menu has a prompt requesting information from the user.

11. A system for aiding user navigation through a voice recognition menu having plural nodes, the system comprising:

a voice recognition response unit interfaced with a telephone network, the voice recognition response unit operable to accept a user request for a menu node submitted as a voice utterance and determine the node of the voice recognition menu associated with the voice utterance;

a confidence engine interfaced with the voice recognition response unit and operable to assign one of plural confidence levels to the determination by the voice recognition response unit of the node associated with the voice utterance;

an interrogate engine interfaced with the voice recognition response unit and operable to command an implicit confirmation for a first confidence level, an explicit confirmation for a second confidence level and a concise interrogative prompt for a third confidence levels; and

a DTMF engine interfaced with the voice recognition response unit, the DTMF engine operable to provide a touch tone menu for selection of a requested node;

wherein the voice recognition response unit further comprises a counter to count user attempts to navigate to a particular node, the counter activating the DTMF engine to provide touch tone navigation after a predetermined number of attempts.

12. The system of claim 11 wherein the implicit confirmation comprises stating the determined node to the user and proceeding to the determined node unless the user refutes the determined node.

13. The system of claim 12 wherein the implicit confirmation further comprises stating the determined node as a query.

14. The system of claim 11 wherein the explicit confirmation comprises stating the determined node to the user as a query requiring an affirmative response before proceeding to the determined node.

15. The system of claim 14 wherein the explicit confirmation further comprises instructions to restate the requested node if the user refutes the determined node.

16. The system of claim 11 wherein the concise interrogative prompt comprises a single word query to restate the requested node.

17. The system of claim 11 wherein the voice recognition response unit detects a predetermined delay by users in responding to requests for input, the voice recognition response unit operable after the predetermined delay to provide a secondary request for input, the secondary request including instructions to input a DTMF response to navigate to a node.

18. The system of claim 11 wherein the DTMF engine selectively activates to offer touch tone menu options in conjunction with voice activated menu options.

19. The system of claim 11, wherein:

the received voice utterance is associated with a particular one of the plural nodes; and

different ones of the plural nodes have different quantitative confidence level parameters such that the assignment of a confidence level to the determination of the node associated with the voice utterance is based at least in part on the quantitative confidence level parameters corresponding with the particular node.

20. A method for voice activated navigation of a menu, the method comprising:

receiving a voice request to navigate to a menu selection;

determining the menu selection by comparing the voice request to a glossary of sample requests;

determining a confidence level for the menu selection determination;

selectively confirming the menu selection determination based on the confidence level;

counting user attempts to state a menu selection; and

automatically providing a DTMF menu to the user after a predetermined number of attempts.

21. The method of claim 20 wherein selectively confirming comprises one of implicit confirmation or explicit confirmation.

22. The method of claim 20 further comprising re-prompting the voice request if the user refutes the menu selection determination.

23. The method of claim 20 wherein determining a confidence level further comprises determining a high, medium or low confidence level, the method further comprising:

querying the user with a concise interrogative prompt for low confidence menu selection determinations.

24. The method of claim 23 wherein the concise interrogative prompt comprises a query having a single word.

25. The method of claim 24 wherein the concise interrogative prompt comprises the query “

Pardon?”

26. The method of claim 20 wherein providing a DTMF menu further comprises offering the user selection by either voice activation or a touch tone input.

27. The method of claim 20, wherein:

the menu includes multiple nodes;

the received voice request is associated with a particular one of the multiple nodes; and

different ones of the multiple nodes have different quantitative confidence level parameters such that the determination of the confidence level for the menu selection determination is based at least in part on the quantitative confidence level parameters corresponding with the particular node.