ChatGPT and Ensemble Learning for Mobile Text Misinformation Identification

Abstract

According to Pew Research Center, eight-in-ten Americans acquire news from digital devices, favoring mobile devices over desktops and laptops. News is therefore spread faster, wider, and easier. However, many of these mobile messages are at risk of being incorrect or even distorted on purpose. This research aims to mitigate this problem by identifying mobile text misinformation to allow mobile users to accurately judge the messages they receive. The proposed method uses various mobile data mining technologies including ChatGPT and several ensemble learning methods (including recurrent neural networks (RNN) and bagging, boosting, stacking, & voting means) to identify mobile misinformation. In addition, sentiment and emotional analyses are discussed in comparison. Experiment results show the ensemble learning methods provide higher accuracy than standalone ChatGPT or RNN model. Nevertheless, the problem, misinformation identification, is intrinsically difficult. Further refinements are needed before it is put into practical use.