Detecting Fake News in Social Media using Natural Language Processing by Fake Polarity Detection

Fake News (FN) has emerged as a significant issue on social media, influencing public perception and rapidly spreading misinformation. The sheer volume of information shared online, primarily via social media networks, is constantly growing. The widespread availability and rapid expansion of data through these networks have made it progressively more challenging to distinguish between genuine and FN. The basic problem addressed in this study is the widespread distribution of FN on social media platforms, which has an important effect on the public's opinion and trust in online information. The explosion of data on these platforms, along with users' predisposition for relying on unconfirmed sources, contributed to the disinformation problem. This study aims to create an efficient FN detection model by leveraging Sentiment Analysis (SA) and Natural Language Processing (NLP) methods. Techniques like Term Frequency-Inverse Document Frequency (TF-IDF), Bag of Words (BoW), and Word2Vec are employed in the process. The proposed system classifies news content into two polarities: positive or negative. The Faker library was utilized to generate realistic synthetic datasets, which enhanced the performance of these NLP techniques. It contributed to achieving better confidence scores, with Word2Vec demonstrating the highest accuracy for detecting both positive and negative polarities. This study contributes to the prevention of misinformation by providing a computational, exact, and user-friendly approach for FN identification, hence strengthening the credibility of social media platforms.