System and method for report level confidence

• US 20060026003A1
• Filed: 07/28/2005
• Published: 02/02/2006
• Est. Priority Date: 07/30/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Lapshina, Larissa, Carus, Alwin B., Lovance, Elizabeth M.
• Granted Patent: US 7,818,175 B2
• Time in Patent Office: Days
• US Class Current: 704/275
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A system and method is disclosed for Report Confidence Modeling (RCM) including automatic adaptive classification of ASR output documents to determine the most efficient document edit workflow to convert dictation into finished output. The RCM according to the present invention may include a mechanism to predict recognition accuracy of a document generated by an ASR engine. Predicted accuracy of the document allows an ASR application to sort recognized documents based on their estimated accuracy or quality and route them appropriately for further processing, editing and/or formatting.

Claims

1. A method of transcribing speech comprising the steps of:

receiving a dictation;

generating a string of recognized text from the dictation;

determining a confidence score for the string of recognized text; and

routing the dictation and the string of recognized text based on the confidence score for final editing.

2. The method of claim 1 wherein the confidence score is determined by modeling.

3. The method of claim 2 wherein the modeling has a threshold component and a scoring strictness component.

4. The method of claim 3 wherein the-threshold component is based on the level of accuracy of the recognized text.

5. The method of claim 3 wherein the scoring strictness component provides an acceptable range of confidence scores for the recognized text.

6. The method of claim 3 wherein the modeling also includes a cost effectiveness component.

7. The method of claim 1 wherein the dictation and recognized text are routed to an adaptation modeling module for final editing.

8. The method of claim 7 wherein automatic speech recognition software performs the editing.

9. The method of claim 1 wherein the dictation and recognized text are routed to a transcriptionist for editing.

10. The method of claim 1 wherein the string of recognized text is generated by automatic speech recognition software.

11. The method of claim 1 wherein the dictation and recognized text also are routed based on predetermined filter settings.

12. The method of claim 1 further comprising the step of providing a report confidence model.

13. The method of claim 12 wherein the confidence score is determined from the report confidence model.

14. The method of claim 12 further comprising the step of customizing the report confidence model based on collected reports, automatic speech recognizer documents, and/or descriptive information corresponding to the corrected reports and automatic speech recognizer documents.

15. A method of transcribing speech comprising the steps of:

receiving a dictation;

generating a string of recognized text from the dictation;

determining a confidence score for the string of recognized text;

routing the dictation and the string of recognized text to an adaptation modeling module or a transcriptionist for final editing based on the confidence score.

16. The method of claim 15 wherein the confidence score is determined to be below a minimum confidence level.

17. The method of claim 16 wherein the dictation and recognized text are routed to the transcriptionist.

18. The method of claim 15 wherein the confidence score is determined to be at or above a minimum confidence level.

19. The method of claim 18 wherein the dictation and recognized text are routed to the adaptation modeling module of automatic speech recognition software.

20. The method of claim 15 wherein the confidence score is determined by modeling.

21. The method of claim 20 wherein the modeling has a threshold component and a scoring strictness component.

22. The method of claim 21 wherein the threshold component is based on the level of accuracy of the recognized text.

23. The method of claim 21 wherein the scoring strictness component provides an acceptable range of confidence scores for the recognized text.

24. The method of claim 21 wherein the modeling also includes a cost effectiveness component.

25. The method of claim 15 wherein the dictation and recognized text also are routed based on predetermined filter settings.

26. The method of claim 15 further comprising the step of providing a report confidence model.

27. The method of claim 26 wherein the confidence score is determined from the report confidence model.

28. The method of claim 26 further comprising the step of customizing the report confidence model based on collected reports, automatic speech recognizer documents, and/or descriptive information corresponding to the corrected reports and automatic speech recognizer documents.

29. A system for transcribing speech from a dictation comprising:

a CPU;

a computer code operatively associated with said CPU, said computer code including;

a first set of instructions configured to generate a string of recognized text from the dictation;

a second set of instructions configured to determine a confidence score for the string of recognized text; and

,

a third set of instructions configured to route the dictation and the recognized text based on the confidence score for final editing.

30. The system of claim 29 wherein the confidence score is determined by modeling.

31. The system of claim 30 wherein the modeling has a threshold component and a scoring strictness component.

32. The system of claim 30 wherein the threshold component is based on the level of accuracy of the recognized text.

33. The system of claim 30 wherein the scoring strictness component provides an acceptable range of confidence scores for the recognized text.

34. The system of claim 30 wherein the modeling also includes a cost effectiveness component.

35. The system of claim 29 wherein the dictation and recognized text are routed to an adaptation modeling module for final editing.

36. The system of claim 35 wherein automatic speech recognition software performs the editing.

37. The system of claim 29 wherein the dictation and recognized text are routed to a transcriptionist for editing.

38. The system of claim 37 wherein the string of recognized text is generated by automatic speech recognition software.