How do we know which physicians will be good candidates for speech recognition?

Successful candidates:  a)  Speak clearly   b)  Speak with good volume   c)  Use consistent terminology   d)  Dictate headings and punctuation   e)  Dictate in calm, quiet locations

What steps will ensure success with speech recognition

1)  Perform ‘Context Tuning’, which means adding site-specific phrases and local terminology and places to the Context, or medical specialty vocabulary, before the go-live date.  This is typically done by the transcription supervisor.

2) Complete ‘Context Adaptation’.  This process updates TA⁺ Powered by SpeechMagic with each dictator’s unique language model.  Context adaptation is accomplished by sending corrected documents back through the ASR engine for learning of each author’s voice patterns.  The TA⁺ system can be set to send corrected transcripts back through the ASR servers, in either the ‘Active’ or ‘Passive’ mode.

3) In preparation for ‘Active’ or live ASR, we recommend that ‘Passive’ ASR take place first.  ‘Passive’ ASR occurs when traditionally-created transcripts are sent through the speech recognition engine for pre-learning.

4) Standardize the text and dictated formats for dates and measurements.  ASR-delivered text will be consistent to the department’s agreed-upon formats.  The greater the dictator consistency with a department’s standardization for dates and dosages, the greater the accuracy in the delivered text.  For example, all dictators would say “February second 2011” and the delivered text would be either ‘2-2-2011’ or ‘Feb. 2, 2011’ or ‘February 2, 2011’, whichever method is preferred by the department.

5) Dictate in a quiet location, so that ASR does not attempt to translate background noises, such as traffic, loud printers or nearby conversations as text.

6) Have physicians alert the system that they wish to insert a Normal or pre-defined text by stating “Insert Normal Chest” instead of just saying “Normal chest” so that the system recognizes the command and responds correctly, instead of continuing to interpret the voice as text.   (Recommended, but not required)

How quickly after dictation will the TA⁺ Powered by SpeechMagic delivered draft text be available?

Text, returned from the TA⁺ Powered by SpeechMagic engines, will be available within a few minutes after dictation. This holds true for onsite license and ASP arrangements. The ‘raw’ text will be allocated to Correctionists for review and editing through the TA^{+ system.}

How does the system improve accuracy over time?

After jobs are edited by a Correctionist with Active ASR or edited by a Transcriptionist with Passive ASR, TA⁺ Powered by SpeechMagic workflow sends the revised version back through the speech recognition servers to continue ConText Adaptation.

What is the ‘Correction Threshold’ field on the Physician User Management screen of TA⁺?

This is the field where the transcription supervisor enters a percentage value related to initial speech recognition accuracy and errors. If the speech rec engine judges the voice file error rate to be lower than the threshold percentage, the job will process through speech recognition. If the voice file error rate is greater than the threshold percentage, the job will be rejected by the speech recognition engine and sent to traditional transcription.

How can I analyze the progress we’re making with speech recognition?

TA⁺ provides several ASR management reports to analyze individual physician performance by work type and over time. Line counting reports show the reduction in traditionally transcribed lines and increase in edited (corrected) lines. Turnaround reports will show the improved turnaround time by work type and by dictator.

Does TA⁺ Powered by SpeechMagic integrate with Arrendale’s e-Sign⁺?

Yes, for sites that use both TA⁺ Powered by SpeechMagic and e-Sign⁺