System and method for report level confidence
First Claim
Patent Images
1. A computer implemented method of transforming a dictation into text comprising steps of:
- receiving said dictation by a computer;
automatically transcribing said dictation into recognized text by automatic speech recognition;
determining a document level confidence score measurement from said recognized text, said document level confidence measurement determination based, at least in part, on word-level recognition accuracy in said automatic transcription and at least one document-level feature affecting how easily the document may be edited or transcribed;
establishing a predetermined confidence measurement threshold value;
submitting said recognized text for editing if said confidence measurement exceeds said predetermined confidence measurement threshold value; and
manually transcribing said dictation into text if said confidence measurement does not exceed said predetermined confidence measurement threshold value.
8 Assignments
0 Petitions
Accused Products
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.
70 Citations
13 Claims
-
1. A computer implemented method of transforming a dictation into text comprising steps of:
-
receiving said dictation by a computer; automatically transcribing said dictation into recognized text by automatic speech recognition; determining a document level confidence score measurement from said recognized text, said document level confidence measurement determination based, at least in part, on word-level recognition accuracy in said automatic transcription and at least one document-level feature affecting how easily the document may be edited or transcribed; establishing a predetermined confidence measurement threshold value; submitting said recognized text for editing if said confidence measurement exceeds said predetermined confidence measurement threshold value; and manually transcribing said dictation into text if said confidence measurement does not exceed said predetermined confidence measurement threshold value. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
Specification