QUESTION: AI create correspondent instead of using existing ones?

[cbirkenbeul]

Hi,
maybe i understood this not right. I can manage which tags are assigned to the documents.
My first batch created a lot of new correspondents instead of using exising ones. Mostly it created the company with full address addiontally.
Is there a way around this or, even better, let paperless-ai use the existing ones.
Thanks.

[mamema]

see issue #37 (comment)

[mamema]

dunno if this here gets trough, but i'm not sure if i understand the solution. The logs are telling "document already has a correspondent" - no change (or similiar worded).
But! Isn't it the case that with the wisdom of paperless-ngx with their learning engine EVERY document gets a correspondent assigned with fresh imported documents? Just that these correspondents are wrong sometimes.....
So, the AI intelligence isn't touching them?
Isn't this contra indictive because a "higher" intelligence gets screwed by a lower one?

So the approach is now what? don't let paperless-ai running over a paperless-ngx with just imported but not manually human proved content?

[clusterzx]

The most people here do want it that way. I changed it just this update. Before it was like you wanted.

But you can just tell paperless to not auto assign a correspondent. Easy fix.

[mamema]

oh, really?, didn't know this.....and i didn't want it another way, i just need to understand whats happening. My impression of the earlier solution was, that (and that is what i hoped for...) used existing correspondents, but perhaps better ones as those already assigned. Instead your solution created new ones which mostly was not better than the already assigned correspondent. But perhaps thats not possible. One guy could dream....