A Study on the Evolution of Public Sentiment Towards Hotspot Events in the Opinion Field of Short Video Platforms-An Example of a Toxic Train Derailment in East Palestine, Ohio

Abstract

Against the background that short video platforms have become the core position of public opinion on public emergencies, this study focuses on the dynamics of public emotion evolution on the TikTok platform of the 2023 Ohio poison train accident. Based on the amplification effect of the algorithmic recommendation mechanism and instant interaction, the study selects 7,199 valid comments during the peak period of communication after the accident (5-27 February), combined with standard deviation polarisation measure and K-means clustering analysis, a three-dimensional dynamic model of "polarity-polarisation-heat" was constructed. The study finds that: firstly, public sentiment shows three stages of evolution - the outbreak period (5-7 February) is dominated by violently fluctuating negative sentiments (42.7% of anger), the diffusion period (8-15 February) is dominated by neutral rationality (31%), but the standard deviation of polarisation increases by 19%; the recession period (16 February) is dominated by decreasing public sentiment heat. In the period of diffusion (8-15 days), neutral rationality was expressed (31%) but the standard deviation of sentiment polarisation increased by 19%. The study suggests that platform mechanisms significantly drive emotional evolution - highly interactive videos reinforce negative emotional expression and opinion polarisation, homogenised 'resonance zones' formed during the high emotion period accelerate the aggregation of extreme emotions, and algorithmic recommendations are more likely to amplify emotional polarisation than traditional social media. This study has important practical significance for optimizing the transparency of platform algorithm and designing emotional intervention nodes in crisis response.