CLAIM CRACKER - A CLAIM DETECTOR AND VALIDATOR

Abstract

The modem digital era produces a large volume of data on an hourly basis around the world including Pakistan. This data contains a significant amount of false, misleading, and fake information and claims. One of the major reasons for the unstable political situation in Pakistan is the spreading of fake claims. These fake claims result in a quick fire around the public causing unfavourable events. Although these fake claims may get exposed as false or invalid claims later, the damage had been done till then. The reason is the manual research for the validation of these claims which takes lot of time to find out the truth. Automated claim detectors and validators can fill this gap. It will result in preventing serious damages by providing the reality of claims in a matter of few minutes. By using advanced natural language processing and machine learning techniques which involve Recurrent Neural Network (RNN), LSTM, Bi-LSTM, CNN and GRU, we have prepared a model trained on past Pakistani political and influential claims and their reality. It is capable of detecting claims in sentences and validating them. To make this model easily accessible to everyone we have made a ReactJS- based modem web application with the Django backend along with the ClaimBuster, arXiv and OpenAI API. It integrates the ClaimCracker model with the interactive user interface of the ClaimCracker web application. This dynamic web application provides a way for the end user to detect and validate claims with just a single button click. The model is capable of providing 93% accuracy for detecting and validating claims based on Pakistani politics and current affairs.