Voice-enabled dialog system

• US 7,869,998 B1
• Filed: 12/19/2002
• Issued: 01/11/2011
• Est. Priority Date: 04/23/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property II LP (AT&T, Inc.)
• Inventors: Di Fabbrizio, Giuseppe, Dutton, Dawn L, Gupta, Narendra K., Hollister, Barbara B., Rahim, Mazin G, Riccardi, Giuseppe, Schapire, Robert Elias, Schroeter, Juergen
• Primary Examiner(s): Hudspeth, David R
• Assistant Examiner(s): Spooner, Lamont M
• Application Number:
  US10/325,295
• Time in Patent Office: 2,945 Days
• Field of Search: 704/1, 704/9, 704/231, 704/251, 707 2- 5
• US Class Current: 704/251
• CPC Class Codes: G10L 13/00 Speech synthesis; Text to speech systems
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A voice-enabled help desk service is disclosed. The service comprises an automatic speech recognition module for recognizing speech from a user, a spoken language understanding module for understanding the output from the automatic speech recognition module, a dialog management module for generating a response to speech from the user, a natural voices text-to-speech synthesis module for synthesizing speech to generate the response to the user, and a frequently asked questions module. The frequently asked questions module handles frequently asked questions from the user by changing voices and providing predetermined prompts to answer the frequently asked question.

Claims

1. A voice-enabled spoken dialog service comprising:

an automatic speech recognition module comprising a general-purpose acoustic model and a domain-specific model, wherein the general-purpose model is used for bootstrapping at initial deployment of the spoken dialog service and the domain-specific model is used to adapt the automatic speech recognition module after deployment, wherein a language corpus for the domain-specific acoustic model is drawn from at least one domain-specific website;

a spoken language understanding module that performs text normalization, entity extraction and semantic classification using a boosting approach that balances human-crafted rules with available data;

a dialog management module that comprises an interpreter, finite state machine engine and an action template;

a text-to-speech synthesis module for synthesizing speech; and

a frequently asked questions module, wherein the spoken dialog service communicates with the user in a first voice, and wherein the frequently asked questions module performs;

receiving an utterance from the user;

determining whether the utterance is a frequently asked question;

if the utterance is a frequently asked question, answering the frequently asked question in a second voice; and

after answering the frequently asked question in the second voice, providing a prompt in the second voice to return the user to the first voice.

2. The voice-enabled spoken dialog service of claim 1, wherein the frequently asked questions module uses a question and answer table having data extracted from previous callers to a spoken dialog system.

3. The voice-enabled spoken dialog service of claim 2, wherein the question and answer table is further partitioned into small subsets, each having a corresponding semantic tag, wherein if the user asks a question that closely matches a question found in the questions and answer table, the answer is automatically passed to the dialog management module with any entities and semantic tags.

4. The voice-enabled spoken dialog service of claim 1, wherein the text-to-speech synthesizer module synthesizes speech using a screenwriting dialog technique where a back story is created for the synthesized voice based on a desired character trait.

5. The voice-enabled spoken dialog service of claim 4, wherein the desired character trait is cheerful.

6. The voice-enabled spoken dialog service of claim 4, wherein the desired character trait is trustworthy.

7. The voice-enabled spoken dialog service of claim 4, wherein the desired character trait is helpful.

8. The voice-enabled spoken dialog service of claim 4, wherein the desired character trait is calm.

9. The voice-enabled spoken dialog service of claim 4, wherein the desired character trait is strong-willed.

10. The voice-enabled spoken dialog service of claim 1, wherein the second voice is of a different gender from the first voice.

11. The voice-enabled spoken dialog service of claim 1, wherein the second voice has different characteristics from the first voice.

12. The voice-enabled spoken dialog service of claim 1, wherein if the frequently asked question is a first frequently asked question asked by the user, the method further comprises:

presenting a first frequently asked question prompt in the second voice.

13. The voice-enabled spoken dialog service of claim 1, wherein the first frequently asked question prompt instructs the user that a virtual agent associated with the second voice handles the frequently asked questions.

14. The voice-enabled spoken dialog service of claim 12, wherein the frequently asked questions module further performs:

after answering the frequently asked question in the second voice, providing a prompt in the second voice to return the user to the first voice.

15. The voice-enabled spoken dialog service of claim 1, wherein the frequently asked questions module further performs:

if the utterance is a frequently asked question, determining whether the frequently asked question is a first frequently asked question by the user;

if the utterance is the first frequently asked question asked by the user;

providing a first frequently asked question introduction prompt using the second voice; and

answering the frequently asked question in the second voice; and

if the utterance is not the first frequently asked question asked by the user;

providing an introductory prompt different from the first frequently asked question introductory prompt in the second voice; and

answering the frequently asked question in the second voice.

16. The voice-enabled spoken dialog service of claim 15, wherein the frequently asked questions module further performs:

providing a prompt in the first voice that a first voice agent associated with the first voice has returned to continue interacting with the user.