Numbered List Detection Explained

Summary

One of the ways that First Draft adds value to basic recognition results is by automatically detecting numbered lists and formatting them in the document. Numbered lists are typically detected by looking for phrases like “number 1” and “number 2” as well as phrases like “number next” and “next number.” If First Draft detects several of these types of phrases close to each other in the document, it will format them into a numbered list and automatically remove the phrases to save editing time.

In order to correctly detect as many numbered lists as possible, First Draft also considers phrases like “number for” and “number to” (which may come up instead of “number 4” or “number 2”). It also takes into account cases where an author skips a number or it is not recognized correctly and attempts to keep the list going despite the missing segment. Tolerances for how many numbers can be missed, how far apart a cluster of number phrases can be, and how long before a list is considered over are all configurable using the Numbered List Detection Sensitivity slider. The slider is located in the Professional Edition section of the FD Tuning page.

Increasing the sensitivity will increase all of the above mentioned tolerances. This will help with Authors who dictate lengthy numbered list items and with recognition results which miss some of the numbers the author dictates. Decreasing the sensitivity will make the numbered list algorithm more conservative when formatting results as lists. This helps in situations where an author may dictate a lot of numbers which are falsely being recognized as lists.

For example, consider the following recognition results:

Plan number 1 continue him on his cholesterol medication. Another flu monitor him for any further side effects, especially dizziness and nausea. Number 3 start him on another hypertension medication to get his blood pressure under control number 4 ….

In this sentence, the author may have coughed as he was saying number 2 and it was incorrectly recognized as "another flu."  This resulted in some space between number 1 and number 3. Setting the numbered list sensitivity to -2 would result in this being formatted into a list, but a setting of -3 would not.

We were originally considering using a number 1 placement tube for this procedure. After consulting with the patient and her family prior to surgery, and after some careful consultation between myself and doctor Smith, we decided to move to a number 3. The risks…

In this case, the author has used number phrases in the middle of a normal description. A sensitivity setting of 5 would format this into a numbered list, but a sensitivity setting of 4 would not. The InfraWare default (0) strikes a balance between low false positives (too many lists) and low false negatives (missed lists) which will be appropriate for most Authors.  When you find the numbered list recognition to missing lists or finding them where they shouldn't be, you can adjust this setting on a per Author basis to improve performance.

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.