Closed quevon24 closed 2 years ago
Looks pretty cool! @flooie can you say more about your process and confidence level?
@mlissner I drafted up a short explainer on how we generated this in issue #34. There is one tricky question about how to handle the roughly 2% that are actually missing opinions.
My confidence level otherwise is pretty high for a task this ... large. I have yet to find any bad results from reviewing +1k generated files.
Very cool and thank you for the write up. It's helpful. I guess it's your call if you think this is done enough. If you looked at 1,000 items and found none that had issues, that's pretty darned good.
The other thing we discussed was tweaking the message in the opinion-less items. That might be worth doing before merging, if it's not too much trouble.
Anyhow, nice work. Pretty cool what's possible these days!
Opinionated fixes for cases that don’t have an opinion defined explicitly inside tag.
This is the solution for issue #2311: https://github.com/freelawproject/courtlistener/issues/2311
This goal was achieved thanks to @flooie and his excellently trained machine learning model