Automatic voice response system using voice recognition means and method of the same

• US 20010010714A1
• Filed: 01/16/2001
• Published: 08/02/2001
• Est. Priority Date: 01/26/2000
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Nemoto, Kazuo
• Granted Patent: US 6,584,180 B2
• Time in Patent Office: Days
• US Class Current: 379/88.01
• CPC Class Codes: H04M 1/642 storing speech in digital form
• CPC Class Codes: H04M 1/663 Preventing unauthorised calls to a telephone set

Abstract

An automatic voice response section transfers voice input from a caller to a voice recognition section, and the voice recognition section generates a voice recognition result. The voice recognition result is synthesized at the voice response section into a voice confirmation message and sent back to the caller for confirmation that the voice input was correctly recognized. When a response by the caller to the confirmation message is negative, the automatic voice response section plays the previously received voice input from the caller as audio to a screener interface. A screener listens to the audio and enters a screener recognition result at the screener interface. The screener recognition result is synthesized into a voice response message by the voice response unit and transmitted to the caller for confirmation. Only if the caller responds negatively to the message based upon the screener recognition result is the call switched to an operator for manual handling of the callers special problems.

Claims

1. Automatic voice interaction apparatus for receiving a voice request from a caller and for transmitting a voice message to the caller comprising:

switching means for receiving a call and for connecting the call to;

automatic voice response means for storing a callers voice utterance and for synthesizing a voice message to the caller;

voice recognition means connected to receive the callers voice utterance for analyzing the voice utterance and for generating a voice recognition result, the voice recognition means connected to send the voice recognition result to the automatic voice response means for synthesis into a voice message to the caller;

screening means for determining when the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance, the screening means including screener interface means connected to receive stored voice utterances and reproduce voiced audio from a stored voice utterance, the screener interface including means for receiving a voice recognition result entered by a screener of the reproduced voiced audio, the screener interface connected to send the entered recognition result to the automatic voice response means for synthesis into a voice message to the caller;

operator interface means by which an operator and the caller directly talk with each other, the screening means controlling the switching means to switch a call from the automatic voice response means to the operator interface means.

2. The automatic voice interaction apparatus of

claim 1

further comprising logic means in the screening means for determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the caller responds negatively to a confirmation voice message synthesized by the voice response section from a recognition result generated by the voice recognition means.

3. The automatic voice interaction apparatus of

claim 1

further comprising:

input means in the screener interface means for receiving input from the screener for controlling the switching means to switch a call from the automatic voice response means to the operator interface means.

4. The automatic voice interaction apparatus of

claim 1

further comprising:

logic means in the screening means for controlling the switching means to switch a call from the automatic voice response means to the operator interface means after determining that the voice recognition means and the screener both have difficulty in recognizing intelligible information in the callers voice utterance when the caller responds negatively to a confirmation voice message synthesized by the voice response section from a recognition result entered by the screener at the screener interface.

5. The automatic voice interaction apparatus of

claim 1

further comprising:

storage means for storing a list of words to be recognized, the voice recognition means selecting a word from proposed words in the list of words corresponding to an intelligible response to a synthesized question message to generate a recognition result.

6. The automatic voice interaction apparatus of

claim 1

further comprising:

logic means in the screening means for determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the caller responds negatively to a confirmation voice message synthesized by the voice response section from a highest probability recognition result generated by the voice recognition means and the caller responds negatively to a confirmation voice message synthesized by the voice response section from a second highest probability recognition result generated by the voice recognition means.

7. The automatic voice interaction apparatus of

claim 1

further comprising:

logic means in the screening means for determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the recognition results generated by the voice recognition means all have low probability of being the intelligible information intended by the caller.

8. Method of automatic voice interaction for receiving a voice request from a caller and for transmitting a voice message to the caller comprising the steps of:

receiving a call at automatic voice response means;

storing a callers voice utterance;

analyzing the callers voice utterance in voice recognition means for intelligible information;

generating a voice recognition result containing the intelligible information;

synthesizing the voice recognition result into a voice message to the caller;

determining when the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance,

reproducing at a screener interface, voiced audio from stored voice utterance;

receiving at the screener interface, a screener'"'"'s recognition result entered by a screener of the reproduced voiced audio;

synthesizing the screener'"'"'s recognition result into a voice message to the caller; and

controlling switching means to switch a call from the automatic voice response means to an operator interface for allowing an operator to speak directly with the caller when the caller responds negatively to the synthesized voice message based upon the screener'"'"'s recognition result.

9. Method of automatic voice interaction of

claim 8

wherein the step of determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance further comprises the step of accepting the caller'"'"'s negative response to a confirmation voice message synthesized by the voice response section from a recognition result generated by the voice recognition means as determinative.

10. Method of

claim 8

further comprising:

receiving input at the screener interface for controlling the switching means to switch a call from the automatic voice response means to the operator interface means.

11. Method of

claim 8

further comprising:

selecting from a stored list of proposed words to be recognized, a word from proposed words in the list of words corresponding to an intelligible response to a synthesized question message to generate a recognition result.

12. Method of

claim 8

wherein the step of determining further comprises:

determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the caller responds negatively to a confirmation voice message synthesized from a highest probability recognition result generated by the voice recognition means and the caller responds negatively to a confirmation voice message synthesized from a second highest probability recognition result generated by the voice recognition means.

13. Method of

claim 8

wherein the step of determining further comprises:

determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the recognition results generated by the voice recognition means all have low probability of being the intelligible information intended by the caller.

14. A computer program product comprising:

a computer useable medium having computer readable program code means embodied therein for operating an automatic voice interaction system which receives a voice request from a caller and transmits a voice message to the caller comprising;

program code means for receiving a call at an automatic voice response system;

program code means for storing a callers voice utterance;

program code means for analyzing the callers voice utterance for intelligible information;

program code means for generating a voice recognition result containing the intelligible information;

program code means for synthesizing the voice recognition result into a voice message to the caller;

program code means for determining when automatic voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance,

program code means for reproducing at a screener interface, voiced audio from stored voice utterance;

program code means for receiving at the screener interface, a screener'"'"'s recognition result entered by a screener of the reproduced voiced audio;

program code means for synthesizing the screener'"'"'s recognition result into a voice message to the caller; and

program code means for controlling switching means to switch a call from the automatic voice response means to an operator interface for allowing an operator to speak directly with the caller when the caller responds negatively to the synthesized voice message based upon the screener'"'"'s recognition result.

15. Computer program product of

claim 14

wherein the program code means for determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance further comprises:

program code means for accepting the caller'"'"'s negative response to a confirmation voice message synthesized by the voice response section from a recognition result generated by the voice recognition means as determinative.

16. Computer program product of

claim 14

further comprising:

program code means for receiving input at the screener interface for controlling the switching means to switch a call from the automatic voice response means to the operator interface means.

17. Computer program product of

claim 14

further comprising:

program code means for selecting from a stored list of proposed words to be recognized, a word from proposed words in the list of words corresponding to an intelligible response to a synthesized question message to generate a recognition result.

18. Computer program product of

claim 14

wherein the program code means for determining further comprises:

determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the caller responds negatively to a confirmation voice message synthesized from a highest probability recognition result generated by the voice recognition means and the caller responds negatively to a confirmation voice message synthesized from a second highest probability recognition result generated by the voice recognition means.

19. Computer program product of

claim 14

wherein the program code means for determining further comprises:

program code means for determining that the voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance when the recognition results generated by the voice recognition means all have low probability of being the intelligible information intended by the caller.

20. A computer program product comprising computer readable program code means for operating an automatic voice interaction system which receives a voice request from a caller and transmits a voice message to the caller comprising:

program code means for receiving a call at an automatic voice response system;

program code means for storing a callers voice utterance;

program code means for analyzing the callers voice utterance for intelligible information;

program code means for generating a voice recognition result containing the intelligible information;

program code means for synthesizing the voice recognition result into a voice message to the caller;

program code means for determining when automatic voice recognition means has difficulty in recognizing intelligible information in the callers voice utterance,

program code means for reproducing at a screener interface, voiced audio from stored voice utterance;

program code means for receiving at the screener interface, a screener'"'"'s recognition result entered by a screener of the reproduced voiced audio;

program code means for synthesizing the screener'"'"'s recognition result into a voice message to the caller; and

program code means for controlling switching means to switch a call from the automatic voice response means to an operator interface for allowing an operator to speak directly with the caller when the caller responds negatively to the synthesized voice message based upon the screener'"'"'s recognition result.

21. An automatic voice response system comprising:

automatic voice response means for receiving a voice from a caller and for transmitting a message to the caller;

voice recognition means for analyzing the voice and for generating a voice recognition result;

screener interface means for reproducing the voice and for entering the voice recognition result by a screener thereof;

operator interface means by which an operator and the caller directly talks with each other; and

line exchanging means for switching a talking with the caller between the automatic voice response means and the operator interface means,

wherein the automatic voice response means transfers the voice of the caller to the voice recognition means, the voice recognition means generates the voice recognition result to transmit it to the automatic voice response means, and the automatic voice response means transmits a confirmation message based on the voice recognition result to the caller;

when a response by the caller to the confirmation message is negative, the automatic voice response means supplies the voice of the caller already received to the screener interface means, the automatic voice response means receives the voice recognition result from the screener interface means to transmit the confirmation message based on the voice recognition result to the caller; and

when the response by the caller to the confirmation message is negative, the line exchanging means switches the talking of the caller from the automatic voice response means to the operator interface means.