Methods and systems for adapting a model for a speech recognition system

• US 9,202,458 B2
• Filed: 10/11/2010
• Issued: 12/01/2015
• Est. Priority Date: 02/04/2005
• Status: Active Grant

• Current Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Original Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Inventors: Braho, Keith P., Pike, Jeffrey P., Pike, Lori A.
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US12/901,870
• Publication Number:
  US 20110029313A1
• Time in Patent Office: 1,877 Days
• Field of Search: 704/251
• US Class Current: 1/1
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/065 Adaptation

Abstract

A method for model adaptation for a speech recognition system includes determining an error rate, corresponding to either recognition of instances of a word or recognition of instances of various words, without using a transcript of words input to the system. The method may further include adjusting an adaptation, of the model for the word or various models for the various words, based on the error rate. Apparatus are disclosed for identifying possible errors made by a speech recognition system without using a transcript of words input to the system. An apparatus for model adaptation for a speech recognition system includes a processor adapted to estimate an error rate, corresponding to either recognition of instances of a word or recognition of instances of various words, without using a transcript of words input to the system. The apparatus may further include a controller adapted to adjust an adaptation of the model for the word or various models for the various words, based on the error rate.

Claims

1. A method for adapting a model for a speech recognition system, comprising:

determining an expected response;

acquiring a speech input using a speech-input device;

generating a hypothesis from the speech input, the hypothesis comprising words, each word having an associated confidence factor;

with an apparatus using at least one hardware-implemented processor, automatically identifying an instance of a word (i) recognized by the system within a certain confidence factor range and (ii) included in a hypothesis generated by the system that matches the expected response as a possible error;

adding the identified possible error to a count of possible errors; and

using the count of possible errors, adjusting the adaption of a model for the word associated with the possible error.

2. The method of claim 1, wherein the range comprises values corresponding to low confidence that the system recognized the instances of the word correctly for an application in which the system is used.

3. The method of claim 1, wherein the range is a predetermined set of values in between a low confidence threshold and a high confidence threshold.

4. The method of claim 3, wherein the low confidence threshold is an adjusted acceptance threshold, which is a value lower than the acceptance threshold.

5. The method of claim 3, wherein the high confidence threshold is a value equal to or higher than an acceptance threshold for accepting the hypothesis generated by the system, in which a confidence value associated with the hypothesis must exceed the acceptance threshold in order for the hypothesis to be accepted.

6. An apparatus for adapting a model for a speech recognition system, comprising:

a speech-input device; and

a processor that is operable to;

(i) determine an expected response,

(ii) acquire a speech input from the speech-input device,

(iii) generate a hypothesis from the speech input, the hypothesis comprising words, each word having an associated confidence factor,

(iv) identify an instance of a word, recognized by the system within a certain confidence factor range and included in the hypothesis generated by the system that matches the expected response, as a possible error;

(v) add the identified possible error to a count of possible errors, and

(vi) use the count of possible errors to adjust the adaption of a model for the word associated with the possible error.

7. The apparatus of claim 6, wherein the range comprises values corresponding to low confidence that the system recognized the instances of the word correctly for an application in which the system is used.

8. The apparatus of claim 6, wherein the range is a predetermined set of values in between a low confidence threshold and a high confidence threshold.

9. The apparatus of claim 8, wherein the high confidence threshold is an acceptance threshold for accepting the hypothesis generated by the system, in which a confidence threshold associated with the hypothesis must exceed the acceptance threshold in order for the hypothesis to be accepted.

10. The apparatus of claim 8, wherein the low confidence threshold is an adjusted acceptance threshold, which is a value lower than the acceptance threshold.

11. The apparatus of claim 8, wherein the high confidence threshold is a value equal to or higher than an acceptance threshold for accepting the hypothesis generated by the system, in which a confidence value associated with the hypothesis must exceed the acceptance threshold in order for the hypothesis to be accepted.