Data-Driven Voice User Interface

• US 20100145710A1
• Filed: 12/08/2008
• Published: 06/10/2010
• Est. Priority Date: 12/08/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Tremblay, Real
• Granted Patent: US 8,165,887 B2
• Time in Patent Office: Days
• US Class Current: 704/275
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A method for developing a voice user interface for a statistical semantic system is described. A set of semantic meanings is defined that reflect semantic classification of a user input dialog. Then, a set of speech dialog prompts is automatically developed from an annotated transcription corpus for directing user inputs to corresponding final semantic meanings. The statistical semantic system may be a call routing application where the semantic meanings are call routing destinations.

Claims

1. A method for developing a voice user interface for a statistical semantic system, the method comprising:

defining a set of semantic meanings that reflect semantic classification of a user input dialog; and

automatically developing from an annotated transcription corpus a set of speech dialog prompts for directing user inputs to corresponding final semantic meanings.

2. A method according to claim 1, wherein the statistical semantic system is a call routing application wherein the semantic meanings are call routing destinations.

3. A method according to claim 1, wherein the speech dialog prompts include one or more confirmation prompts for confirming semantic meaning in a given user input.

4. A method according to claim 1, wherein the speech dialog prompts include one or more back-off prompts for reprompting a given user input.

5. A method according to claim 1, wherein the speech dialog prompts include one or more disambiguation prompts for intermediate concept classification of a given user input.

6. A method according to claim 1, wherein automatically developing a set of speech dialog prompts includes:

developing a set of data archetypes from the annotated transcription corpus wherein each data archetype represents one or more annotated transcription samples that are determined to be semantically similar;

developing a set of semantic meaning archetypes by associating one or more of the data archetypes with each semantic meaning; and

dynamically developing a dialog prompt for a current user input based on a determination of a best matching semantic meaning archetype for the current user input.

7. A method according to claim 1, wherein automatically developing a set of speech dialog prompts includes:

extracting from the annotated transcription corpus one or more semantic meaning informative keywords associated with one or more of the semantic meanings; and

dynamically developing a dialog prompt for a current user input based on the extracted keywords.

8. A method according to claim 1, wherein the annotated transcription corpus is a training corpus used for training one or more semantic classification processes associated with directing user inputs.

9. A computer program product in a computer readable storage medium for execution on at least one processor of a method of developing a voice user interface for a statistical semantic system, the computer program product having instructions for execution on the at least one processor comprising;

program code for defining a set of semantic meanings that reflect semantic classification of a user input dialog; and

program code for automatically developing from an annotated transcription corpus a set of speech dialog prompts for directing user inputs to corresponding final semantic meanings.

10. A product according to claim 9, wherein the statistical semantic system is a call routing application wherein the semantic meanings are call routing destinations.

11. A product according to claim 9, wherein the speech dialog prompts include one or more confirmation prompts for confirming semantic meaning in a given user input.

12. A product according to claim 9, wherein the speech dialog prompts include one or more back-off prompts for reprompting a given user input.

13. A product according to claim 9, wherein the speech dialog prompts include one or more disambiguation prompts for intermediate concept classification of a given user input.

14. A product according to claim 9, wherein the program code for automatically developing a set of speech dialog prompts includes:

program code for developing a set of data archetypes from the annotated transcription corpus wherein each data archetype represents one or more annotated transcription samples that are determined to be semantically similar;

program code for developing a set of semantic meaning archetypes by associating one or more of the data archetypes with each semantic meaning; and

program code for dynamically developing a dialog prompt for a current user input based on a determination of a best matching semantic meaning archetype for the current user input.

15. A product according to claim 9, wherein the program code for automatically developing a set of speech dialog prompts includes:

program code for developing a set of annotated transcription samples from the annotated transcription corpus by determining for selected annotated transcription samples one or more characteristic keywords associated with one or more of the semantic meanings; and

program code for dynamically developing a dialog prompt for a current user input based on a determination of a best matching annotated transcription sample for the current user input.

16. A product according to claim 9, wherein the annotated transcription corpus is a training corpus used for training one or more semantic classification processes associated with directing user inputs.

17. A system for developing a voice user interface for a statistical semantic system, the system comprising;

means for defining a set of semantic meanings that reflect semantic classification of a user input dialog; and

means for automatically developing from an annotated transcription corpus a set of speech dialog prompts for directing user inputs to corresponding final semantic meanings.

18. A system according to claim 17, wherein the statistical semantic system is a call routing application wherein the semantic meanings are call routing destinations.

19. A system according to claim 17, wherein the speech dialog prompts include one or more confirmation prompts for confirming semantic meaning in a given user input.

20. A system according to claim 17, wherein the speech dialog prompts include one or more back-off prompts for reprompting a given user input.

21. A system according to claim 17, wherein the speech dialog prompts include one or more disambiguation prompts for intermediate concept classification of a given user input.

22. A system according to claim 17, wherein the means for automatically developing a set of speech dialog prompts includes:

means for developing a set of data archetypes from the annotated transcription corpus wherein each data archetype represents one or more annotated transcription samples that are determined to be semantically similar;

means for developing a set of semantic meaning archetypes by associating one or more of the data archetypes with each semantic meaning; and

means for dynamically developing a dialog prompt for a current user input based on a determination of a best matching semantic meaning archetype for the current user input.

23. A system according to claim 17, wherein the means for automatically developing a set of speech dialog prompts includes:

means for developing a set of annotated transcription samples from the annotated transcription corpus by determining for selected annotated transcription samples one or more characteristic keywords associated with one or more of the semantic meanings; and

means for dynamically developing a dialog prompt for a current user input based on a determination of a best matching annotated transcription sample for the current user input.

24. A system according to claim 17, wherein the annotated transcription corpus is a training corpus used for training one or more semantic classification processes associated with directing user inputs.