Misinformation Analysis During Covid-19 Pandemic

Abstract

Online diffusion of misinformation has gained extreme attention in the research from past few years. Moreover, during ongoing Covid-19 pandemic, the proliferation of misinformation became more prominent. In this paper, a comparison of two feature engineering techniques, namely term frequency–inverse document frequency (tf-idf) and word embeddings (doc2vec), is done over different machine learning classifiers. A Web scraper is developed for fact-checking Web site, Snopes.com, to extract labeled news related to Covid-19. Although the size of dataset is less, the body content under headlines contains large amount of text. Therefore, the model works well with both the feature engineering techniques and machine learning algorithms. Apparently, we obtained best accuracy of 95.38% with tf-idf on decision tree and same accuracy of 90.77% using doc2vec on support vector machine and logistic regression machine learning classifier.