A Natural Language Processing (NLP) tool to detect and rebut vaccine-related misinformation using a fine-tuned BERT model with OCR support. Future enhancements will include a Retrieval-Augmented Generation (RAG) pipeline that provides real-time, evidence-based rebuttals.
- Text Classification: BERT-based binary classifier (
misinfovs.factual) - OCR Detection: Extract and classify text from images (e.g., memes)
- Future RAG Pipeline:
- Semantic search over a curated vaccine knowledge base
- Contextual rebuttal generation with citations
Check out the live demo hosted on Hugging Face Spaces:
https://huggingface.co/spaces/Heishigan/vaccine-misinfo-detector
git clone https://github.com/Heishigan/vaccine-misinfo-detector.git
cd vaccine-misinfo-detector
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py
Or use Gradio:
gradio app.py
- Input: Text or image containing a vaccine-related claim
- Classification: BERT model outputs 0 (misinfo) or 1 (factual)
- (Upcoming): If flagged as misinformation, the RAG pipeline will launch a knowledge retrieval + LLM-generated rebuttal
Developed by:
- Heishigan
- Xinrui Wan
- Kaif Ahmed
- Siqi Xiang
Built with:
- Hugging Face Transformers
- EasyOCR
- Gradio
- SciVerify
- World Health Organization (WHO)
- Centers for Disease Control and Prevention (CDC)
- ✅ Deploy to Hugging Face Spaces
- ✅ Add OCR image support
- ⬜ Implement RAG pipeline (retriever + generator)
- ⬜ Add citation metadata to rebuttals
- ⬜ Evaluate on external datasets