Abstract:
Fake news basically means fabricating a story without verifiable information, source or quote of the story. Check That! 2021 Task-3 has two subtasks, subtask 3a and 3b. We participated in Subtask 3a, which is a multi-class fake news classification problem. The goal was to determine whether the main claim of a news article is true, partially true, false or other. We were provided with a dataset of news articles by the organizers which consists of news articles, their titles and the rating of the article. We took advantage of TF-IDF vectorization and proposed an Extreme Gradient Boosting algorithm for our best classification model. The approaches were very interpretative with a highest classification accuracy of 0.57 and highest f1-macro score of 0.54 on the given dataset. We also tried other classification models and got varying results which are simple Logistic Regression Classifiers, Passive Aggressive Classifiers and Random Forest Classifiers
