How to apply this guide
Check the speaker and add a time reference before attributing 2004 and 2009 refer to different events in the transcript.
Start with
Use the approved interview cut and its release notes to find a clear introduction or a longer statement from the relevant speaker. Label that voice in a working copy first. Check who actually says “I moved to the town in 2004, but the shop did not open until 2009.”, including any person being quoted.
Check before you finish
Review the approved interview transcript as the recipient will see it. A label should clearly apply to the correct line. For a short excerpt, include enough context to show whether the speaker was giving their own view or reporting someone else’s.
Evidence to keep: Record the expected sentence, actual result, microphone, language, model and app build before changing settings. This page’s key limit is: Check the speaker and add a time reference before attributing 2004 and 2009 refer to different events in the transcript.
Start here
Speech recognition and speaker attribution are separate checks. With a recorded community interview, a person’s name beside an accurate phrase can still be wrong when speakers hand off quickly or talk over each other.
Establish a voice reference
Use the approved interview cut and its release notes to find a clear introduction or a longer statement from the relevant speaker. Label that voice in a working copy first. Check who actually says “I moved to the town in 2004, but the shop did not open until 2009.”, including any person being quoted.
Mark the uncertain boundary
Keep the approximate playback time where the speaker changes. If needed, leave a temporary “speaker unclear” marker and ask the interview editor to confirm attribution. Keep whether the guest approved the web excerpt open if the answer depends on that unclear turn.
Verify the attributed line on export
Review the approved interview transcript as the recipient will see it. A label should clearly apply to the correct line. For a short excerpt, include enough context to show whether the speaker was giving their own view or reporting someone else’s.




