What is it about?
Propasafe is a browser extension that is based on offline inference. That means no data leaves user's machine to infer whether the sentence the user reads is propaganda or non propaganda. The tool is also modular in design.
Featured Image
Photo by Flipboard on Unsplash
Why is it important?
The tool helps users read news articles more cautiously by providing color-coded highlights of potential propaganda techniques with reasonable accuracy. It serves as an alternative to LLM-based inference tools, which often incur API credit costs. By being free and open-source, the tool reduces the cost of inference while maintaining transparency. Its modular design also allows advanced users to replace the pre-trained BERT model with more efficient alternatives.
Perspectives
While other tools rely on LLM-based inference, this tool fills the gap for users who require offline propaganda detection to avoid API credit costs while maintaining reasonable detection performance. It also addresses privacy concerns by ensuring that no user data is sent to external LLMs or developers. The tool is open source and modular, allowing users to easily update it by replacing the fine-tuned model with a more efficient one.
Vivek Sharma
City University of New York Graduate School and University Center
Read the Original
This page is a summary of: Propasafe: A BERT-based Offline Tool for Propaganda Detection, May 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3701716.3715195.
You can read the full text:
Resources
Contributors
The following have contributed to this page
