Skip to main content

How to Add or Remove Labels for Model Training

This guide helps you update your dataset to add or remove categories and ensure the AI learns the right patterns to train your model.

Kristie Maclyman avatar
Written by Kristie Maclyman
Updated over a week ago

1. How to Add a Label

To teach the model a new category (e.g., “feature request” or “login failure”), follow these steps:

➕ Steps:

  • Add the new label to the Categories column.

  • Add the corresponding parent label (e.g., Functionality, Access Issues) to the Parent Categories column.

  • Use a pipe (|) to separate multiple entries.

  • Order matters: each child category should follow its parent in left-to-right order.

📌 Example:

Before

Child Categories: payment error | delay 
Parent Categories: Billing | Delivery

After

Child Categories: login failure | payment error 
Parent Categories: Access Issues | Billing

⚠️ Note: The model reads categories from left to right, so the order of parents and children must match.

2. How to Remove a Label

To remove a label (e.g., "delay") from your dataset:

➖ Steps:

  • Delete the label from the Categories column.

  • If its corresponding parent (e.g., "Delivery") is no longer needed, remove it from Parent Categories as well.

Summary: Preparing Your Data

  • Keep only the rows you want to modify.

  • To train the model effectively:

    • Provide at least 15 examples per new category.

    • Include at least 5 varied examples per new category to help the model understand the concept across different contexts.

  • Add new category and parent labels in matching order.

  • Keep existing labels unless you want to update or remove them.

  • Once your file is ready, upload the updated dataset to your dashboard—the model will retrain accordingly.

Did this answer your question?