Comment on page

Datasaur Predictive Labeling

Generate labels by providing specific tasks or questions that you want the system to answer
Datasaur predictive Labeling will allow you to optimize prediction performance when preparing labeled data to reduce the annotator's time consumption. Label less, predict more!
Here's a quick guide on using Datasaur Predictive Labeling for Row Based project:
  1. 1.
    Enable Datasaur Predictive Labeling extension in your selected project by navigating to the extension settings.
    Datasaur Predictive Labeling in extension settings
  2. 2.
    After enabling it, you can see the input and output fields. You can select the Input Columns as the context and the Target Field as the column for the predicted answer.
    Datasaur Predictive Labeling Extension
  3. 3.
    Click "Save Configuration" and let the magic start! If your project already has some labeled data, the system will show you predictions. If not, you need to label a minimum of 5 data for each category of the answer. For example, if you have two categories: POSITIVE and NEGATIVE, you need to label at least 5 data as POSITIVE and 5 data as NEGATIVE.
  4. 4.
    After the predictions are displayed, you can accept or reject the labels based on your convenience and review them accordingly.
    Accept and Reject the predictions
Last modified 3mo ago