Method, apparatus, and computer program product for one-step correction of voice interaction

• US 20060247913A1
• Filed: 04/29/2005
• Published: 11/02/2006
• Est. Priority Date: 04/29/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Pieraccini, Roberto, Huerta, Juan Manuel
• Granted Patent: US 7,720,684 B2
• Time in Patent Office: Days
• US Class Current: 704/1
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A one-step correction mechanism for voice interaction is provided. Correction of a previous state is enabled simultaneously with recognition in a current or subsequent state. An application is decomposed into a set of tasks. Each task is associated with the collection of one piece of information. Each task may be in a different state. At any point during the interaction, while a task/state pair is active, the dialog manager may enable multiple other task/state pairs to be active in latent fashion. The application developer may then use those facilities or resources to the active task/state and the latent task/state pairs depending on contextual condition of the interaction state of the application.

Claims

1. A method, in a data processing system, for correction of voice interaction, the method comprising:

providing an active grammar for an active state;

receiving a first piece of information for the active state spoken by a user;

prompting the user for a second piece of information for a second state, wherein the first state becomes a latent state and the second state becomes the active state;

providing a latent grammar for a set of one or more latent states including the first state;

determining whether correction to the first piece of information is received; and

responsive to correction to the first piece of information being received, using the latent grammar to correct the first piece of information.

2. The method of claim 1, wherein prompting the user for the second piece of information includes echoing the first piece of information.

3. The method of claim 1, further comprising:

outputting an explicit confirmation of the corrected first piece of information.

4. The method of claim 1, further comprising:

responsive to a determination that the first state has been corrected a predetermined maximum number of times, invoking a failure handler.

5. The method of claim 4, wherein the predetermined maximum number of times is specified in a policies file.

6. The method of claim 1, further comprising:

responsive to receiving the second piece of information for the active state spoken by a user, updating a next active state and updating the set of one or more latent states to include the second state.

7. The method of claim 5, further comprising:

prompting the user for a third piece of information for the next active state.

8. The method of claim 6, wherein prompting the user for the third piece of information includes echoing the second piece of information.

9. The method of claim 1, wherein the set of one or more latent states is specified in a policies file.

10. The method of claim 1, wherein determining whether correction to the first piece of information is received includes determining whether user input complies with the latent grammar.

11. An apparatus, in a data processing system, for correction of voice interaction, the apparatus comprising:

a speech recognizer;

a prompt player;

a dialog manager;

an active grammar for an active state; and

a latent grammar for a set of one or more latent states including the first state,

wherein the dialog manager receives from the speech recognizer a first piece of information for the active state spoken by a user, prompts the user via the prompt player for a second piece of information for a second state, wherein the first state becomes a latent state and the second state becomes the active state, determines whether correction to the first piece of information is received, and uses the latent grammar to correct the first piece of information responsive to correction to the first piece of information being received.

12. The apparatus of claim 11, wherein the dialog manager prompts the user for the second piece of information and echoes the first piece of information in the same interaction.

13. The apparatus of claim 11, wherein the dialog manager outputs an explicit confirmation of the corrected first piece of information via the prompt player.

14. The apparatus of claim 11, wherein the dialog manager invokes a failure handler responsive to a determination that the first state has been corrected a predetermined maximum number of times.

15. The apparatus of claim 14, wherein the predetermined maximum number of times is specified in a policies file.

16. The apparatus of claim 11, wherein the dialog manager updates a next active state and updates the set of one or more latent states to include the second state responsive to receiving the second piece of information for the active state spoken by a user.

17. The apparatus of claim 16, wherein the dialog manager prompts the user for a third piece of information for the next active state via the prompt player.

18. The apparatus of claim 17, wherein the dialog manager prompts the user for the third piece of information and echoes the second piece of information in the same interaction.

19. The apparatus of claim 11, wherein the set of one or more latent states is specified in a policies file.

20. The apparatus of claim 11, wherein the dialog manager determines whether correction to the first piece of information is received by determining whether user input complies with the latent grammar.

21. A computer program product, in a computer readable medium, for correction of voice interaction, the computer program product comprising:

instructions for providing an active grammar for an active state;

instructions for receiving a first piece of information for the active state spoken by a user;

instructions for prompting the user for a second piece of information for a second state, wherein the first state becomes a latent state and the second state becomes the active state;

instructions for providing a latent grammar for a set of one or more latent states including the first state;

instructions for determining whether correction to the first piece of information is received; and

instructions responsive to correction to the first piece of information being received, for using the latent grammar to correct the first piece of information.

22. The computer program product of claim 21, wherein prompting the user for the second piece of information includes echoing the first piece of information.

23. The computer program product of claim 21, further comprising:

instructions for outputting an explicit confirmation of the corrected first piece of information.

24. The computer program product of claim 21, further comprising:

instructions for invoking a failure handler responsive to a determination that the first state has been corrected a predetermined maximum number of times.

25. The computer program product of claim 24, wherein the predetermined maximum number of times is specified in a policies file.

26. The computer program product of claim 21, further comprising:

instructions for updating a next active state and updating the set of one or more latent states to include the second state responsive to receiving the second piece of information for the active state spoken by a user.

27. The computer program product of claim 26, further comprising:

instructions for prompting the user for a third piece of information for the next active state.

28. The computer program product of claim 27, wherein prompting the user for the third piece of information includes echoing the second piece of information.

29. The computer program product of claim 21, wherein the set of one or more latent states is specified in a policies file.

30. The computer program product of claim 21, wherein the instructions for determining whether correction to the first piece of information is received include instructions for determining whether user input complies with the latent grammar.