cdp-frontend icon indicating copy to clipboard operation
cdp-frontend copied to clipboard

Change transcript selection from highest confidence to most recent

Open evamaxfield opened this issue 2 years ago • 2 comments

I have realized that the confidence of our old "webvtt transcripts" have a higher confidence score than our new whisper transcripts which is just.... not correct in many cases.

This is a larger discussion about the deprecation of "confidence" in our metadata but for now the fast solution is to update our transcript selection to most recently created.

evamaxfield avatar Mar 04 '23 16:03 evamaxfield

@evamaxfield cdpf256 I think this is it but I'm not sure where the transcript ordering is on the actual frontend site is. Can you link me to it and then I'll test if it's correct or not.

jonlin14 avatar Mar 08 '23 19:03 jonlin14

@jonlin14 I think that might be the only spot that needs to be changed haha!

You should be able to try it locally by updating the local config: https://github.com/CouncilDataProject/cdp-frontend/blob/main/src/index-react.tsx#L7

to be the seattle production config: https://github.com/CouncilDataProject/seattle/blob/main/web/src/index.jsx#L7

The events that have been reprocessed with Whisper will be the older ones first (2020 to 2021 so far).

Toggle it back and forth and hopefully see the transcript update to the better version?

Honestly, it might be good to update the example frontend config in this repo to point at the seattle production data rather than the seattle-staging data too.

evamaxfield avatar Mar 09 '23 00:03 evamaxfield