nico-timm/toxic-comment-detection

Toxic Comment Detection

This web application uses a fine-tuned DistilBERT model to classify comments across six toxicity categories:

• —Toxic
• —Severe Toxic
• —Obscene
• —Threat
• —Insult
• —Identity Hate

How to Use

1. 1.Enter a comment in the text box.
2. 2.Click Analyze Comment.
3. 3.Review the predicted probability, classification threshold, and detection result for each category.

The model uses validation-selected thresholds for each toxicity category rather than a universal 0.50 threshold.

Project

Source Code: https://github.com/Nico-Timmerman/toxic-comment-classification

Fine-Tuned Model: https://huggingface.co/nico-timm/toxic-comment-distilbert