New updates and improvements to Caravel.

October 30, 2020

Training feedback

This week we've taken a small but essential step towards letting the user know when the system is revising its predictions.

When a classifier is created or trained on data, it immediately starts to improve its predictions. This process can take a few minutes, depending on the amount of data. The problem is without any feedback loop; it can leave the user unsure if the new predictions are ready. When a circle loader appears next to a classifier's name, you'll know it's working on improving its predictions.

Other improvements & fixes

• Public release of redaction classifiers
• Bug Fix Fixed unresponsive worker in cosine-similarity scoring
• UI copy changes to our filter controls

October 12, 2020