Factify is an intelligent fake news detector that uses BERT (DistilBERT) embeddings along with a machine learning classifier to evaluate whether a news article is real or fake, based on its language patterns.
This tool does not verify facts or sources. It analyzes textual writing style and patterns to assist in identifying potential misinformation. Always verify news with credible sources.
- ✍️ Paste any news article or paragraph and get instant analysis
- 🧠 Uses BERT-based semantic understanding for contextual language analysis
- 📊 Displays confidence scores for predictions
- 🎨 Clean and modern UI powered by Streamlit
- 🔒 Fully local execution — no external API calls or data sharing
- Frontend/UI: Streamlit (custom themed)
- Modeling: HuggingFace Transformers (DistilBERT), Scikit-learn
- Classifier: Logistic Regression
- Data Handling: Pandas, Torch
- Persistence: Joblib (for saving/loading the model)
The dataset files (Fake.csv, True.csv) have been moved into a compressed archive to reduce repository size.
➡️ You’ll find them inside the data.zip file in the root directory.
Shreeya Soma