Automatic speech recognition (ASR) processing using confidence measures

• US 5,566,272 A
• Filed: 10/27/1993
• Issued: 10/15/1996
• Est. Priority Date: 10/27/1993
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Brems, Douglas J., Schoeffler, Max S.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): ONKA, THOMAS
• Application Number:
  US08/144,065
• Time in Patent Office: 1,084 Days
• Field of Search: 381/110, 381/41, 381/43, 379/59, 395/2.4
• US Class Current: 704/231
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

The user interface in an automatic speech recognition (ASR) system is dynamically controlled, based upon the level of confidence in the results of the ASR process. In one embodiment, the system is arranged to distinguish error prone ASR interpretations from those likely to be correct, using a degree of confidence in the output of the ASR system determined as a function of the difference between the confidence in the "first choice" selected by the ASR system and the confidence in the "second choice" selected by the ASR system. In this embodiment, the user interface is arranged so that the explicit verification steps taken by the system as a result of uncertain information is different from the action taken when the confidence is high. In addition, different treatment can be provided based upon the "consequences" of misinterpretation as well as the historical performance of the system with respect to the specific user whose speech is being processed. In another embodiment, after an ASR system interprets an utterance, the confidence in the interpretation is assessed, and three different interactions with the user may then occur.

Claims

1. A system for adapting the user interface in systems that:

accept speech input and perform automatic speech recognition (ASR), comprising

means for receiving .an utterance;

means for processing said utterance using ASR to generate an interpretation of said utterance and to determine a level of confidence in said interpretation; and

means for selectively adapting the verification of said interpretation requested from the user as a function of said confidence level,

wherein said system further includes

means for storing, for each user of said system, information representing a success measure computed as a function of previous uses of said system, and

means for retrieving information from said storing means and for adapting said user interface as a function of the value of said success measure.

2. The invention defined in claim 1 wherein said success measure includes the previous success rate for said each user of said system.

3. The invention defined in claim 1 wherein said success measure includes previous confidence values associated with ASR interpretations for said each user.

4. The invention defined in claim 1 wherein said system is arranged to compare said success measure to a user dependent threshold.

5. An automatic speech recognition system comprising

means for generating an interpretation of a user'"'"'s utterance and a confidence value for said interpretation, and

user interface means operative in response to the magnitude of said confidence value, for (a) requesting said user to verify said interpretation prior to accepting said interpretation as an accurate representation of said utterance, or (b) accepting said interpretation as an accurate representation of said utterance without verification.

wherein said system further includes means for storing information indicative of the previous success of said system in interpreting utterances of said user, and

means for responsive to said stored information for controlling said user interface means.

6. A method of adapting the user interface in systems that accept speech input and perform automatic speech recognition (ASR), comprising the steps of

receiving an utterance;

processing said utterance using ASR to generate an interpretation of said utterance and to determine a level of confidence in said interpretation; and

selectively adapting the verification of said interpretation requested from the user as a function of said confidence level,

storing, for each user of said system, information representing a success measure computed as a function of previous uses of said system, and

retrieving information and altering the user interface as a function of the value of said success measure.

7. The method defined in claim 6 wherein said success measure includes the previous success rate for said each user of said method.

8. The method defined in claim 6 wherein said success measure includes previous confidence values associated with ASR interpretations for said each user.

9. The method defined in claim 6 wherein said method further includes comparing said success measure to a user dependent threshold.

10. A method for performing automatic speech recognition system comprising the steps of

generating an interpretation of a user'"'"'s utterance and a confidence value for said interpretation, and

adapting the operation or a user interface in response to the magnitude of said confidence value, by (a) requesting said user to verify said interpretation prior to accepting said interpretation as an accurate representation of said utterance, (b) accepting said interpretation as an accurate representation of said utterance without verification, or (c) rejecting said interpretation and requesting said user to provide a new utterance,

wherein said method further includes the steps of storing information indicative of the previous success of said system in interpreting utterances of said user, and

adapting said user interface in response to said stored information.