First Draft Phrases Explained

Overview

First Draft Phrases are the various things an Author could say when he or she indicates that he or she is starting a new section/heading of the template. For instance, if you are adding a region for the Medication section of the document, you may want to add “Medications,” “Current Medications,” and “Present Medications” as phrases. When First Draft recognizes any of those phrases in the speech recognition output, it will begin redirecting the text to where the field is located.

Summary

• Linearity:  For a given heading there could be many matches throughout the document, or even many matches clustered on the same sentence.  The algorithm seeks to pick the set of matches that has the highest total confidence, as long as it satisfies the linearity requirement (order of headings on the template).  For example, if there is a set of really high confidence matches in the reverse order of the template, we would not pick these in most cases because it is so opposite from how the template is structured.  Picking those would usually be a false positive.
• Phrase Length:  When you have an FD Region which has multiple phrases, the heading detection will choose just one of those phrases to recognize when it is doing processing.  Many times, there are different phrases that match different parts of the document, so heading detection has to decide which one to use.  Usually, it will pick the longest phrase in the list that has a good match. 
  • A 10-word phrase will be chosen over a 5 word phrase if all things are equal (The main exception is if the longer phrase would cause some other region to not recognize any phrases).
  • The longer the phrase, the more likely the match
• Partial Matching: First Draft supports partial matching of a phrase.  This means that if a word or two is wrong in a long phrase, it will still match. 
  • The sensitivity can be adjusted by the heading sensitivity setting.  See KB567 for details.
  • For each of these full or partial matches, the heading detection algorithm assigns a number of confidence indicators to the match.
• Performance Impact:  Performance can be impacted in the following circumstances:
  • Vague/common phrases such as "of" or "a" or "then" by themselves
  • The same phrase used in multiple regions throughout the document
  • Multiple dictators using the same template when author 1 uses a phrase in the middle of a region and Author 2 uses the same phrase to indicate the start of a new region
• Multiple Phrase Limits:  There is no limit to the number of phrases which can be added to a single region.  Adding multiple options for the same phrase can help.  For example:
  • Add multiple options for the same phrase:  He/she, his/her, with punctuation and without.  Punctuation must have a space before and after.