Analyzing Online Fake News Using Latent Semantic Analysis: Case of USA Election Campaign

Abstract

Recent studies have indicated that fake news is always produced to manipulate readers and that it spreads very fast and brings great damage to human society through social media. From the available literature, most studies focused on fake news detection and identification and fake news sentiment analysis using machine learning or deep learning techniques. However, relatively few researchers have paid attention to fake news analysis. This is especially true for fake political news. Unlike other published works which built fake news detection models from computer scientists’ viewpoints, this study aims to develop an effective method that combines natural language processing (NLP) and latent semantic analysis (LSA) using singular value decomposition (SVD) techniques to help social scientists to analyze fake news for discovering the exact elements. In addition, the authors analyze the characteristics of true news and fake news. A real case from the USA election campaign in 2016 is employed to demonstrate the effectiveness of our methods. The experimental results could give useful suggestions to future researchers to distinguish fake news. This study finds the five concepts extracted from LSA and that they are representative of political fake news during the election.