Detecting and regulating sentiment reversal and polarization in online communities

Sentiment reversals and polarizations can disrupt the harmony within a legitimate and peaceful online communication environment. To fill the research gaps, this paper introduces detection methods grounded in catastrophe theory and proposes two innovative regulatory strategies: reversal control strategy (RCS) and polarization control strategy (PCS). Experiments and empirical analysis are conducted on a self-built dataset encompassing approximately 50,000 user groups from Baidu Tieba. In the detection phase, the stochastic catastrophe model achieves an $R^2$ of 0.57, a reversal index of 0.18 and a polarization index of 0.29, indicating the existence of sentiment reversal and polarization. In the regulation phase, RCS outperforms control groups by up to 53% and PCS outperforms control groups by up to 63%. Our empirical analysis reveals two insights. Firstly, an excessive regulation intensity does not proportionally increase benefits but instead diminishes the effectiveness of regulation. Secondly, strategies aim to preventing sentiment reversals can lead to sentiment polarizations and vice versa. This study holds theoretical and practical significance for the decision-making of online communities’ regulation, and also contributes to the management application of catastrophe theory.