Label Error Detection

Double Check your labels automatically using Label Error Detection

Overview

Datasaur's Label Error Detection feature, automating the identification of inaccuracies in datasets to enhance data integrity and model performance. Manual data review is time-consuming and prone to errors. With automated error detection, efficiency is boosted by pinpointing errors, focus is enhanced by applying specific error thresholds, and accuracy is improved with intelligent label suggestions.

Label Error Detection

Use Case

1. Create Project: Follow the guide here to create a row-based project. Here’s what the data looks like below.

   

   Empty Row Based Project
2. Enable Label-Error Detection: Click 'Manage extension' icon on the right bar and toggle on the Label Error Detection feature.

   

   Manage Extension
3. Labeling the data: Label your data using ML-assisted labeling, predictive labeling, or manual labeling.

   

   Labeling Process
4. Incorporate Label Error Detection: Select the target column in the “Target Column” dropdown. In this case, the target column is “Review” column. After that, select the appropriate target question in the “Target Question" dropdown. Finally, click "Find label errors," and voilà!

   

   Empty Label Error Detection
5. Review the labels: Click “View all” and review the appropriate labels! You can revise them individually or click “Reject all” or “Accept all”.

   

   Label Error Detection Result

From the use case above, users will be able to double-check and get additional perspectives from a dedicated model for the user's project. For further details, please visit the Assisted Review - Label Error Detection.