AUTOMATED FACT-CHECKING IN HINDI: HARNESSING AI FOR INFORMATION VERIFICATION AND HARM DETECTION

Abstract

This dissertation evaluates the use of automated fact-checking models specifically designed for the Hindi Language, with the help of artificial intelligence(AI). As we know nowadays false news spreads like fire among common people, hence it is crucial to prioritize the verification of news and information. Although we know that there are various fact-checking projects available on the internet they mainly focus on the English language, hence leading to limited focus on other widely spoken languages such as Hindi. This study aims to overcome this situation by proposing a new methodology for automatically verifying facts, specifically designed for Hindi Language. Using AI and NLP approaches, we create a system that can analyze and confirm the truthfulness of written statements in Hindi. The creation of this system entails gathering and organizing two separate datasets obtained from well-known Hindi news sources, specifically Aaj Tak and Alt News. The datasets are carefully marked with annotations to aid in the use of supervised learning methods for fact-checking. We evaluate the system’s efficacy in effectively detecting disinformation and verifying factual claims in Hindi-language content through thorough experimentation and assessment.