code-kern-ai
[BUG] - Filter by weak supervision confidence in 2 tasks
Describe the bug I have one classification and one extraction task, for which I have weak supervision labels for both. Now, when I want to filter for low confidence records of task A, task B scores are also considered. This makes it much harder for me to find low-quality records of task A.
To Reproduce I will provide a sample project as a zip file which you can import as a local snapshot. Generally, the steps are:
- Create a project with two labeling tasks
- Create weakly supervised labels for both
- Try filtering for the confidence score; you will see that this filter is always applied on both tasks
Expected behavior Only filter the confidence of the selected task (i.e. the one which's drawer is open)
Screenshots
Desktop (please complete the following information):
- OS: MacOs
- Browser: Chrome
- Version: 1.2.0
Additional context -
This seems to be already fixed - the provided project doesn't yield the same results (less items filtered & scrolling testing with static slice + db query didn't show anything with higher confidence than 0.2959740681299307)
Best quess: Fixed by the confidence issue https://github.com/code-kern-ai/refinery/issues/99 - missing grouping in prepared query
PR that fixed it: https://github.com/code-kern-ai/refinery-gateway/pull/65
