The issue of spread of misinformation on social media platforms has brought a huge paradigm shift in social, political and psychological aspects of life. Current research work focuses on detection of misinformation by analysing the content and how it propagates on social media platforms. In this research, effectiveness of sentiment score as a feature for classification of misinformation is studied. A hypothesis is laid on the grounds that statements expressed with malicious intention are always subject to human perceptions and biases which are not neutral. This leads to a significant feature for classification i.e. identifying and extracting sentiment of the text. The approach proposed makes use of VADER corpus for computing sentiment score of information to be verified. An implicit substantiation of the proposed approach is made by comparing the result of sentiment inclusive feature to that of text-only features for classification. The experiment is performed on state-of-the-art twitter dataset PHEME. The results show that on adding sentiment score to the feature list, good improvement in accuracy is obtained of classifier for credibility assessment.
Volume 12 | 07-Special Issue
Pages: 1738-1745
DOI: 10.5373/JARDCS/V12SP7/20202283