Get Free Synopsis PDF
We will prepare a detailed synopsis for your college submission. Includes problem statement, objectives, DFD, ER diagram, and more.
A news credibility checker that classifies articles and social media posts as real or fake using fine-tuned transformer models. Users paste a link or text and get a credibility score with the phrases that influenced the decision.
Problem Statement
Misinformation spreads faster than fact-checkers can respond. Manual verification does not scale, and keyword-based filters miss context, sarcasm, and reworded claims.
Objectives
Fine-tune a BERT-family model on labelled news datasets
Build a text and URL submission interface
Explain predictions by highlighting influential phrases
Compare transformer accuracy against classical ML baselines
Modules
Data Pipeline
Scrapes article text from URLs, cleans it, and tokenizes it for the model.
Classification Model
Fine-tuned transformer that outputs a real/fake probability.
Explainability Layer
Attention and LIME based highlighting of the phrases behind each prediction.
Web Interface
Flask app for submitting content and viewing credibility reports and history.
Expected Output
A web app where a user submits a news article and receives a credibility score, a real/fake label, and highlighted evidence phrases.
Future Scope
Multilingual support for Indian languages, a browser extension, image and video misinformation checks, and source reputation scoring.