A Literature Review On Multilingual Fake News and Scam Detection for Regional Languages Using Natural Language Processing

Sharing news, messages and links online now takes only a few seconds. The same speed helps false stories, misleading posts, scam messages and harmful links travel across social networks and chat apps. Researchers have tried to catch such content with machine learning, deep learning, transformer-based models, multimodal techniques and automatic fact-checking. However, much of this effort has been built around English and depends on very large datasets, so it does not carry over easily to Indian regional languages like Hindi and Marathi. In this paper we look at earlier work on detecting fake news, processing text in many languages, combining text with images, checking facts automatically, explaining model decisions and spotting suspicious web links. We also consider how people actually write in regional languages, including mixed-language messages, words typed in Roman script and the many formats in which content is shared. From these studies we note several weak points: few regional-language resources, results that are hard to interpret, limited support for different input types, little source checking and no single tool that brings threat detection together. To address these points we put forward RegionalGuard, a framework that examines digital content written in English, Hindi, Marathi and Hinglish. It brings together language detection, message checking, text extraction from screenshots, link inspection, news verification, review of social-media posts and forwarded messages, risk scoring and results written in simple words. The tool is meant mainly for people who use regional languages in their daily communication.