METHOD AND SYSTEM FOR AUTOMATIC TRANSCRIPTION PRIORITIZATION

• US 20120166193A1
• Filed: 01/19/2012
• Published: 06/28/2012
• 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
• Granted Patent: US 8,407,050 B2
• Time in Patent Office: Days
• 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 transcription for validating speech recognition performance comprising:

logging information generated from a speech recognition task of one or more spoken utterances;

categorizing said information into categories; and

prioritizing a transcription of said information based on a category;

wherein said speech recognition task associates a spoken utterance with at least one recognized result, and said prioritizing identifies spoken utterances having a transcription priority in view of said speech recognized result.

2. The method of claim 1, further comprising automatically presenting a transcription of said spoken utterance as a modifiable text, wherein said transcription initially defaults to a text representation of said recognition result, wherein the transcription may be one of a correct or incorrect interpretation of said spoken utterance.

3. The method of claim 1, further comprising presenting said categories in a visual display.

4. The method of claim 1, wherein said information includes at least one of a spoken utterance, a recognized result, a confidence score, and an N-best match, wherein said speech recognition task produces said information during a processing of spoken utterance.

5. The method of claim 4, wherein said categorizing includes at least one of a category for spoken utterances, a category for recognized results, a category for confidence scores, and a category for an N-best match.

6. The method of claim 3, wherein said prioritizing is based on an accuracy rating, wherein said accuracy rating is a weighting of said confidence score by said N-best matches.

7. The method of claim 6, further comprising:

lowering said accuracy rating when said N-best matches have neighbor confidence scores approximating said confidence score.

8. The method of claim 6, further comprising:

adding an overall accuracy evaluation that includes at least one of a phonetic analysis of N-best results, an acoustic confusability of a grammar, and a correlation difference with other spoken utterances producing the same said recognized result.

9. The method of claim 6, wherein said prioritizing identifies information for transcription based on an accept and reject criterion, wherein

an accuracy rating above a threshold identifies validated results;

and

an accuracy rating below a threshold identifies invalided results.

10. The method of claim 9, wherein said information can be identified using a color code.

11. The method of claim 3, further comprising detecting a clicking action in said visual display to automatically fill transcriptions.

12. A visual toolkit for prioritizing speech transcription for validating speech recognition performance comprising:

a logger for capturing information from a speech recognition system during a processing one or more spoken utterances;

a processor for determining an accuracy rating of said information; and

a visual display for categorizing said information into categories, and for prioritizing a transcription of said information based on one said accuracy rating,

wherein said speech recognition system associates a spoken utterance with at least one recognized result, and said prioritizing identifies spoken utterances having a transcription priority in view of said recognized result.

13. The visual toolkit of claim 12, wherein said information includes at least one of a spoken utterance, a recognized result, a confidence score, and an N-best match.

14. The visual toolkit of claim 13, wherein said visual display categorizes said information into:

a spoken utterance category having a play button for audibly presenting said spoken utterance;

a recognized result category having a textbox entry for presenting said recognized result;

a confidence score category identifying a numeric ranking of said recognized result; and

a N-best match category having a pop-up box for presenting a textbox of neighbor recognition results,

wherein a recognition result is a text representation of a recognized spoken utterance that may be one of a correct or incorrect interpretation of said spoken utterance.

15. The visual toolkit of claim 13, further comprising:

a transcription category having a modifiable textbox entry with a text entry initially corresponding to a text of said recognized result; and

an accept button for validating a transcription of said recognized result.

16. The visual toolkit of claim 13, wherein said categories are automatically ranked by said accuracy rating in an ordered priority for transcription.

17. The visual toolkit of claim 16, wherein said ordered priority is identifiable by a color code associated with said transcription priority.

18. A method for prioritizing speech transcription for validating speech recognition performance comprising:

importing log information from a speech recognition system from a recognition task of one or more spoken utterances, wherein said log information includes the spoken utterance, a recognized result, a confidence score, and an N-best match,

presenting said information into categories in a visual display; and

automatically filling a transcription category with a text representation of said recognized result.

wherein said speech recognition system associates a spoken utterance with at least one recognized result.

19. The method of claim 18, further comprising:

determining an accuracy rating for determining a transcription priority, wherein said accuracy rating is a weighting of said confidence score by said N-best matches.

20. The method of claim 19, further comprising:

automatically prioritizing said information in said display in view of said accuracy rating, wherein said prioritizing identifies spoken utterances having a transcription priority in view of said recognized result.

21. (canceled)