-
Notifications
You must be signed in to change notification settings - Fork 80
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[IX] - Statuses revamp #7085
Comments
A note on the behavior and updates of the stats panel:
|
We are removing some styles to drive towards a cleaner/simpler "stats & filters" drawer. (these styles will propagate in the future towards all of the existing and future drawers/sidepanels, including the most important ones in library and entity v2) As discussed with @txau we are removing the "others" title and some percentage values. |
Just to keep info here: |
An addition to this. After a conversation with @gabriel-piles we are going to remove the constraint of having selection rectangles to be considered labeled data for property types "text", "number" and "date". So now they can be used the same as selects, meaning that, if they have a value or they are labeled as empty, then they are considered labeled data. If they are just empty string without explicit "label as empty" then it is not labeled data. This goes along the line of what @konzz was pointing out for the match/mismatch. If it is labeled (has a value or explicit empty) it goes into the match/mismatch stats. |
Putting it all together from: #6979 #6977 #7060 #7066
Filters should look like:
LABELED (X%) ............... amount (Accuracy somewhere)
NON LABELED (X%) ............................ amount
Match ............. amount
Mismatch .............. amount
(OTHERS)
Obsolete ...............amount
Error.......................amount
We need to give support for "labeled as empty". Uwazi needs to send these empty markers along with the rest of the training data. If a property has no value, but it also does not have the empty marker, we do not send it to training.
Statuses are display on the table as:
Models will control what comes as "segment", which can contain either the context or some other message, to be display in the column in the table. Ie. if there was an error, that row will contain the status "Error" along with the context "Couldn't read PDF file" or whatever the model decides. We don't need to run any logic on the rendering of the context, just print whatever the model is sending, if it is a blank string, so be it.
The text was updated successfully, but these errors were encountered: