Method and system for automatic transcription prioritization

• US 8,407,050 B2
• Filed: 01/19/2012
• Issued: 03/26/2013
• Est. Priority Date: 04/11/2006
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Kobal, Jeffrey S., Dhanakshirur, Girish
• Primary Examiner(s): Smits, Talivaldis Ivars
• Assistant Examiner(s): ROBERTS, SHAUN A
• Application Number:
  US13/354,142
• Publication Number:
  US 20120166193A1
• Time in Patent Office: 432 Days
• Field of Search: 704/231, 704/235, 704/251
• US Class Current: 704/235
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A visual toolkit for prioritizing speech transcription is provided. The toolkit can include a logger (102) for capturing information from a speech recognition system, a processor (104) for determining an accuracy rating of the information, and a visual display (106) for categorizing the information and prioritizing a transcription of the information based on the accuracy rating. The prioritizing identifies spoken utterances having a transcription priority in view of the recognized result. The visual display can include a transcription category (156) having a modifiable textbox entry with a text entry initially corresponding to a text of the recognized result, and an accept button (157) for validating a transcription of the recognized result. The categories can be automatically ranked by the accuracy rating in an ordered priority for increasing an efficiency of transcription.

Claims

1. A method for prioritizing speech recognition results to be evaluated by a human transcriptionist, the method comprising acts of:

associating a first accuracy rating with a first recognized phrase, the first recognized phrase being a portion of a first recognized text generated at least in part by performing speech recognition processing on one or more first spoken inputs;

associating a second accuracy rating with the second recognized phrase, the second recognized phrase being a portion of a second recognized text generated at least in part by performing speech recognition processing on one or more second spoken inputs; and

presenting the first and second recognized phrases to the human transcriptionist for evaluation, wherein;

the first recognized phrase is presented in such a manner as to be outside a context of the first recognized text and dissociated from at least one other portion of the first recognized text,

the second recognized phrase is presented in such a manner as to be outside a context of the second recognized text and dissociated from at least one other portion of the second recognized text, and

at least one visual indication is provided to reflect a priority between the first and second recognized phrases, the priority being based at least in part on ranking the first and second accuracy ratings.

2. The method of claim 1, wherein the act of associating the first accuracy rating comprises acts of:

determining an initial accuracy value for the first recognized phrase, the initial accuracy value being a confidence score output by an automatic speech recognition process and being associated with a confidence that the first recognized phrase is an accurate recognition of the one or more first spoken inputs; and

adjusting the initial accuracy value to determine the first accuracy rating, wherein the act of adjusting the initial accuracy value is based at least in part on confidence scores associated with candidate phrases in an N-best set of matching candidate phrases for the first recognized phrase.

3. The method of claim 2, wherein the act of adjusting the initial accuracy value comprises acts of:

calculating, for each one of the candidate phrases in the N-best set of matching candidate phrases, a numeric distance between the confidence score associated with the one of the candidate phrases and the confidence score associated with the first recognized phrase; and

calculating a weighted sum of the numeric distances for the candidate phrases in the N-best set of matching candidate phrases.

4. The method of claim 1, wherein the first accuracy rating is higher than the second accuracy rating, and wherein the at least one visual indication comprises presenting the first and second recognized phrases in an order indicative of the first recognized phrase having higher priority than the second recognized phrase.

5. The method of claim 1, wherein the at least one visual indication comprises:

presenting the first recognized phrase in a first color selected based at least in part on the first accuracy rating associated with the first recognized phrase; and

presenting the first recognized phrase in a second color selected based at least in part on the second accuracy rating associated with the second recognized phrase, the second color being different from the first color.

6. The method of claim 1, wherein the one or more first spoken inputs are different from the one or more second spoken inputs.

7. The method of claim 1, further comprising acts of:

presenting at least one first user interface element which, when activated, causes at least a portion of a first audio file to be audibly rendered to the human transcriptionist, the first audio file comprising the one or more first spoken inputs; and

presenting at least one second user interface element which, when activated, causes at least a portion of a second audio file to be audibly rendered to the human transcriptionist, the second audio file being different from the first audio file and comprising the one or more second spoken inputs.

8. The method of claim 1, further comprising acts of:

selecting, based at least in part on the first accuracy rating, the first recognized phrase for presentation to the human transcriptionist; and

selecting, based at least in part on the second accuracy rating, the second recognized phrase for presentation to the human transcriptionist.

9. A system for prioritizing speech recognition results to be evaluated by a human transcriptionist, the system comprising at least one processor programmed to:

associate a first accuracy rating with a first recognized phrase, the first recognized phrase being a portion of a first recognized text generated at least in part by performing speech recognition processing on one or more first spoken inputs;

associate a second accuracy rating with the second recognized phrase, the second recognized phrase being a portion of a second recognized text generated at least in part by performing speech recognition processing on one or more second spoken inputs; and

present the first and second recognized phrases to the human transcriptionist for evaluation, wherein;

the first recognized phrase is presented in such a manner as to be outside a context of the first recognized text and dissociated from at least one other portion of the first recognized text,

the second recognized phrase is presented in such a manner as to be outside a context of the second recognized text and dissociated from at least one other portion of the second recognized text, and

at least one visual indication is provided to reflect a priority between the first and second recognized phrases, the priority being based at least in part on ranking the first and second accuracy ratings.

10. The system of claim 9, wherein the at least one processor is programmed to associate the first accuracy rating at least in part by:

determining an initial accuracy value for the first recognized phrase, the initial accuracy value being a confidence score output by an automatic speech recognition process and being associated with a confidence that the first recognized phrase is an accurate recognition of the one or more first spoken inputs; and

adjusting the initial accuracy value to determine the first accuracy rating, wherein the act of adjusting the initial accuracy value is based at least in part on confidence scores associated with candidate phrases in an N-best set of matching candidate phrases for the first recognized phrase.

11. The system of claim 10, wherein the at least one processor is programmed to adjust the initial accuracy value at least in part by:

calculating, for each one of the candidate phrases in the N-best set of matching candidate phrases, a numeric distance between the confidence score associated with the one of the candidate phrases and the confidence score associated with the first recognized phrase; and

calculating a weighted sum of the numeric distances for the candidate phrases in the N-best set of matching candidate phrases.

12. The system of claim 9, wherein the first accuracy rating is higher than the second accuracy rating, and wherein the at least one visual indication comprises presenting the first and second recognized phrases in an order indicative of the first recognized phrase having higher priority than the second recognized phrase.

13. The system of claim 9, wherein the at least one visual indication comprises:

presenting the first recognized phrase in a first color selected based at least in part on the first accuracy rating associated with the first recognized phrase; and

presenting the first recognized phrase in a second color selected based at least in part on the second accuracy rating associated with the second recognized phrase, the second color being different from the first color.

14. The system of claim 9, wherein the one or more first spoken inputs are different from the one or more second spoken inputs.

15. The system of claim 9, wherein the at least one processor is further programmed to:

present at least one first user interface element which, when activated, causes at least a portion of a first audio file to be audibly rendered to the human transcriptionist, the first audio file comprising the one or more first spoken inputs; and

present at least one second user interface element which, when activated, causes at least a portion of a second audio file to be audibly rendered to the human transcriptionist, the second audio file being different from the first audio file and comprising the one or more second spoken inputs.

16. The system of claim 9, wherein the at least one processor is further programmed to:

select, based at least in part on the first accuracy rating, the first recognized phrase for presentation to the human transcriptionist; and

select, based at least in part on the second accuracy rating, the second recognized phrase for presentation to the human transcriptionist.

17. At least one computer-readable storage device having encoded thereon instructions that, when executed by at least one processor, perform a method for prioritizing speech recognition results to be evaluated by a human transcriptionist, the method comprising acts of:

associating a first accuracy rating with a first recognized phrase, the first recognized phrase being a portion of a first recognized text generated at least in part by performing speech recognition processing on one or more first spoken inputs;

associating a second accuracy rating with the second recognized phrase, the second recognized phrase being a portion of a second recognized text generated at least in part by performing speech recognition processing on one or more second spoken inputs; and

presenting the first and second recognized phrases to the human transcriptionist for evaluation, wherein;

the first recognized phrase is presented in such a manner as to be outside a context of the first recognized text and dissociated from at least one other portion of the first recognized text,

the second recognized phrase is presented in such a manner as to be outside a context of the second recognized text and dissociated from at least one other portion of the second recognized text, and

at least one visual indication is provided to reflect a priority between the first and second recognized phrases, the priority being based at least in part on ranking the first and second accuracy ratings.

18. The at least one computer-readable storage device of claim 17, wherein the act of associating the first accuracy rating comprises acts of:

determining an initial accuracy value for the first recognized phrase, the initial accuracy value being a confidence score output by an automatic speech recognition process and being associated with a confidence that the first recognized phrase is an accurate recognition of the one or more first spoken inputs; and

adjusting the initial accuracy value to determine the first accuracy rating, wherein the act of adjusting the initial accuracy value is based at least in part on confidence scores associated with candidate phrases in an N-best set of matching candidate phrases for the first recognized phrase.

19. The at least one computer-readable storage device of claim 18, wherein the act of adjusting the initial accuracy value comprises acts of:

calculating, for each one of the candidate phrases in the N-best set of matching candidate phrases, a numeric distance between the confidence score associated with the one of the candidate phrases and the confidence score associated with the first recognized phrase; and

calculating a weighted sum of the numeric distances for the candidate phrases in the N-best set of matching candidate phrases.

20. The at least one computer-readable storage device of claim 17, wherein the first accuracy rating is higher than the second accuracy rating, and wherein the at least one visual indication comprises presenting the first and second recognized phrases in an order indicative of the first recognized phrase having higher priority than the second recognized phrase.

21. The at least one computer-readable storage device of claim 17, wherein the at least one visual indication comprises:

presenting the first recognized phrase in a first color selected based at least in part on the first accuracy rating associated with the first recognized phrase; and

presenting the first recognized phrase in a second color selected based at least in part on the second accuracy rating associated with the second recognized phrase, the second color being different from the first color.

22. The at least one computer-readable storage device of claim 17, wherein the one or more first spoken inputs are different from the one or more second spoken inputs.

23. The at least one computer-readable storage device of claim 17, wherein the method further comprises acts of:

presenting at least one first user interface element which, when activated, causes at least a portion of a first audio file to be audibly rendered to the human transcriptionist, the first audio file comprising the one or more first spoken inputs; and

presenting at least one second user interface element which, when activated, causes at least a portion of a second audio file to be audibly rendered to the human transcriptionist, the second audio file being different from the first audio file and comprising the one or more second spoken inputs.

24. The at least one computer-readable storage device of claim 17, wherein the method further comprises acts of:

selecting, based at least in part on the first accuracy rating, the first recognized phrase for presentation to the human transcriptionist; and

selecting, based at least in part on the second accuracy rating, the second recognized phrase for presentation to the human transcriptionist.