Handle incomplete or incorrect speaker attribution
Understand how the scribe decides who said what, and correct the note when it gets a speaker wrong.
Purpose
The scribe labels each line by who it thinks said it, and the note is written from those labels. This guide explains the labels and what to check when a speaker is wrong or missing.
Before you start
the scribe recognises only your voice, from your voice enrolment. Patients aren't enrolled, so the
patient is whoever isn't you.
there's no button to change who said a line. You correct the draft note instead.
Steps
Step 1: Enrol your voice first
Doenrol your voice before your first scribed consultation.
Sowithout it, Keel can't tell which voice is yours, and the lines stay unresolved.
Step 2: Read the speaker labels
You'll seeyour lines on the right and the patient's on the left, each with its role.
You'll seea centred line marked RESOLVING VOICE while Keel hasn't decided who spoke. If only one voice has
been heard so far, it reads SPEAKER 1.
labels can change during the visit as Keel recognises your voice.
Step 3: Check the draft for who said what
Dowhen you review the draft, check anything the patient "reported" that someone else said, and anything said
in the visit that's missing. Edit the note before you sign it.
Sothe note follows the transcript's labels, so a wrong label can put words in the patient's mouth.
Result
Sothe signed note says who said what correctly, even if the transcript didn't.
If it doesn't work
- If you see: Your lines stay RESOLVING VOICE all visitEnrol your voice (step 1). If you're enrolled, check it isn't marked Re-enrol needed.
- If you see: A relative's or nurse's words are on the patient's sideCorrect the note in review (step 3)
- If you see: Something said in the room isn't in the transcriptAdd it to the note by hand (step 3)
- If you see: After a crash: This recording is missing a speakerNo note can be written from it. Recover from interrupted recording
- Working: Lines are on the right sidesDone
Limits
the scribe separates two voices: yours and one other. Once it knows your voice, everyone else is
labelled as the patient.
when three or more people talk, the quietest voice can be left out of the transcript or merged
with another speaker. For example, a quiet patient whose relative does most of the talking.
a separate label for a third person, such as a relative, is not available.
Checked against the Keel product source and in the live product on 2026-10-11. No step on this page was observed. Live transcript labels and the recovery screen were not exercised.