language:
- id
license: mit
library_name: transformers
base_model:
- microsoft/deberta-v3-xsmall
pipeline_tag: text-classification
Anti-Judol DeBERTa: Online Gambling Text Classification Model
Model Description
This model is a fine-tuned version of microsoft/deberta-v3-xsmall specifically designed for detecting and classifying online gambling-related content in Indonesian text. The model helps identify potentially harmful gambling content to support digital safety initiatives and content moderation efforts. It can distinguish between gambling-related and non-gambling text with high accuracy, making it valuable for automated content filtering systems.
Intended Uses
- Content moderation for social media platforms
- Automated detection of gambling advertisements
- Educational content filtering
- Digital safety applications
- Research on harmful content detection
Data
The model was trained on a curated dataset of Indonesian text samples containing both gambling-related and non-gambling content. The dataset includes various types of gambling content such as online casino advertisements, sports betting promotions, lottery schemes, and gambling-related discussions, balanced with legitimate content from news, social media, and educational sources.
Training
| Step | Training Loss | Validation Loss | F1 |
|---|---|---|---|
| 50 | 0.676000 | 0.598374 | 0.806867 |
| 100 | 0.439700 | 0.275495 | 0.911765 |
| 150 | 0.240300 | 0.161314 | 0.950226 |
| 200 | 0.135600 | 0.130217 | 0.966767 |
| 250 | 0.120000 | 0.115214 | 0.971168 |
| 300 | 0.098600 | 0.114574 | 0.975758 |
| 350 | 0.071900 | 0.111686 | 0.978788 |
Ethical Considerations
This model is developed to support digital safety and harm reduction efforts. It should be used responsibly and in compliance with laws and regulations.