More Speaker Recognition Issues

I just had a very frustrating experience trying to generate an accurate meeting summary. There were three participants in the meeting. One of them came into the meeting part way through. Neither of the other two participants were people were in my participant list already.

I tagged the new participant, added his name. It looked good. I scrolled down to where the third person came in. He was listed as the second participant. I thought, no problem, I’ll add the third participant’s name and manually tag him. Not ideal but doable.

Nope, not doable. When I added the third participant’s name, it deleted the second participant’s name not only from the transcript but also from the participant list. A person who had previously been added in another meeting was a choice.

No matter what I did, I could not figure out a way to convince the AI that there were three participants. It insisted that there were only two and it couldn’t correct any of the misattributions. It said that I had 26+ years of institutional experience but in fact, it’s the third participant, the one it refused to recognize as existing, who has that experience.

Is there no way to make it recognize there were three people in the room?

1 Like

Hey @megsaint we are pushing something out to deeply edit the transcript its currently under testing and will be live by the week. So that should solve your issue

Essentially you could take the segment where the third participant came in and create a new speaker there. Which would give the AI more context to generate a summary

2 Likes

Thanks for the status update. I will try to stop pointlessly arguing with the AI.

More of an FYI.

After some further analysis, I think part of my problem may have been due to where the Pocket was placed in the office. It was off to one side next to me, whereas both of the other participants were on the other side of the desk. The one who got missed was the farthest away. On a more detailed review of the transcript, I think some of what he said got missed. He’s fairly soft-spoken.

This doesn’t resolve the issue of being able to re-label. However, it may inform how I place the device in future meetings.

Looking forward to future updates!

Hey @mhirrrr! Will this update allow us to correct the speaker without it affecting other conversations?

I’ve run into issues where it is auto tagging “John” in meeting B, for example, but it’s actually “Chris” in that meeting. So if I change the speaker in meeting B to reflect Chris, it then changes my other recording with John in meeting A to reflect Chris, as well.

Basically in its current state: the more recordings, the greater risk of confusion on who is speaking.

4 Likes

Yes. This is the same issue. I am having as well.

Hey @aimee23 yes we have apps under review and are hoping this will be live soon!

I have experienced this same thing very frustrating