A Systematic Review of Text Mining Techniques for Fake News Detection

In today’s digital environment, fake news spreads rapidly, undermining trust and confidence. This study investigates using text-mining techniques to detect and combat fake news. Through a detailed analysis of fifteen journals, we examine the effectiveness of textual analysis in distinguishing truth from fakery. By examining mechanisms and outcomes, we aim to inform the development of more reliable detection methods, empower stakeholders to fight misinformation in the digital age, and support accurate information. This study examines text mining for fake news detection. The aim is to use text mining to detect fake news to prevent its spread and to detect the source of fake news to curb its veracity. The study adopts literature review techniques, in which fifteen relevant journals were reviewed. This helps to collect the needed data to analyze fake news detection. Many of the reviewed journals used real and fake news datasets, whereas in the same articles, some percentages are labeled as fake and the rest as real. Also, some studies use 80% of their dataset for training, while the remaining 20% is for testing. Also, some journals used 49.9% of their dataset for training and the remaining 50.1% for testing. Different authors deployed many variables; however, the central aim was to define fake news models’ accuracy, precision, and F1 scores. Many of the models are highly accurate to the tune of greater than 80% for fake news detection. Thus, using text mining analytics outperforms the traditional use of journalists, linguists, or media experts to evaluate information credibility. It was found that the strength of text mining analytics lies in its scalability, multimodal analysis, real-time detection, and interpretability. Also, some identified inherent limitations, including semantic complexity, data quality, feature selection, and algorithm biases, were found, and they are the reasons this study calls for multimodal fusion and the development of interpretable models for transparent explanation in future studies.