Label Error Detection

Automatically detect and review potential label errors in your dataset.

Overview

Label error detection helps identify potentially incorrect labels in your dataset to improve data quality and model performance.

The feature automatically detects possible labeling inconsistencies, helping users review data more efficiently and validate labels with additional model-based suggestions.

Label error detection

Quick guide

1. Create a row labeling project.

   
2. Click the gear icon in the extension panel on the right to open the Manage extensions dialog, then enable the Label error detection feature.

   
3. Label your data manually or using ML-assisted labeling or Predictive labeling.

   
4. In the Label error detection extension, select the target column to use as context and the target question, then click Find label errors.

   
5. Click View all and review the detection. You can revise them individually or click Reject all or Accept all.

   

For more information, see Assisted Review - Label Error Detection.