Intelligent Classifier for Identify Reliable On-Demand Messages

Abstract

Accurately extracting useful messages from bodies of information is important. This work proposes an intelligent system, called AI@nti-Fake system, to categories social news and determine whether it is true or false. The news is preprocessed using a Natural Language Processing technique. The text sentiment analysis in the on-demand message is analyzed to identify the fake news. A dataset from the International Workshop on Semantic Evaluation is used in this study. The on-demand message is related to the public’s attention, and the analyzed text sentiment is identified as positive, neutral or negative. The accuracies of the proposed AI@nti-Fake system in the training stage and the real data test can reach 90% and 80%, respectively. The F1-Score of the proposed approach and two others methods are 78.50, 64.84 and 64.59, respectively. The results of the analysis reveal that the F1-Score of our approach can get better performance in classifying on-demand messages and detecting disinformation. The proposed AI@nti-Fake system, which is based on social media analysis and the judgment of sentiment may have applications in business.