Heading Detection Sensitivity Explained

Overview

During the First Draft speech recognition process, headings for each section of the report  are detected automatically based on the phrases specified in the First Draft Regions defined in the template using the ITE and by taking into consideration a setting known as the Heading Detection Sensitivity.  Adjustments to this sensitivity setting are performed in the Professional section of the First Draft Tuning tab in the IMC available for Administrators of the Professional and Expert Editions.  It is available for the entire Account or on a more granular basis for each specific Author:

By increasing the sensitivity of heading detection, you are increasing its tolerance in two areas. The first area is partial phrase matching. For example, suppose we have the heading “History of Present Illness” which we are trying to find. Sometimes speech recognition might only hear part of a word, such as “Ill” instead of “Illness.” With sensitivity at a moderate level, the phrase “History of Present Ill” would still be recognized as matching the heading we were looking for. Partial phrase matching also includes cases where speech recognition completely misses a certain word or the author does not even dictate it. For instance, if speech recognition only recognizes “History of Present,” and heading detection sensitivity is high enough, then that could still be considered a match for “History of Present Illness.” If sensitivity is very low, however, only the exact phrase will be considered a match. Below are examples of some possible speech recognition outputs, and the sensitivity that would be required to match them to “History of Present Illness”

• “History of Present Illness” – any sensitivity level
• “Hist of Present Illness” – minimum of -1 sensitivity level
• “Present Illness” – minimum of 0 sensitivity level
• “His of Ill” – minimum of 2 sensitivity level

When you want more headings to be recognized, consider first whether or not you should add additional phrases to the First Draft Region in the template using the ITE. If an author says “preop diagnosis” instead of “preoperative diagnosis,” then it will be easy to add that phrase to the First Draft Region in the ITE and avoid any changes to heading sensitivity.

A word of caution: turning the sensitivity up higher can sometimes have unexpected consequences. For instance, in our above example, “His” is a partial match on “History.” If the sensitivity level is at 4, then just this one partial match would be considered enough to match on the whole heading. The InfraWare default (0) strikes a balance between low false positives and low false negatives for Authors with “A” and “B” quality results, so for most Authors no tuning will be needed.

The second (and less influential) area the heading detection sensitivity touches is out-of-order dictation. First Draft uses the template to figure out the order in which it should hear the headings dictated. Sometimes, however, an author may dictate the headings out of order from the template. A simple transposition (such as saying “social history” before “family history” when the template is ordered with “family history” first) is common and tolerated even at low sensitivity levels. An author may, however, say “plan” when he or she is dictating the HPI section, but he or she obviously is not intending to start that area of the report. Only a very high sensitivity setting would recognize headings which were dictated 4 or 5 sections out of place like that.

Notes:

• Once this setting is changed and saved it will not modify existing drafts for jobs already on the platform.  This setting change would be applied to new drafts dictated after you make the change.