Understanding Sentiment Polarities and Emotion Categories of People on Public Incidents With the Relation to Government Policies

Public incidents necessitate prompt proactive measures by the government and pertinent departments, posing substantial challenges to emergency management capabilities. With the advancement of internet technologies, social media platforms have played a pivotal role in shaping the landscape of public incidents, progressively emerging as primary conduits for authentic expression and sentiment sharing among individuals. The sentiment polarities and emotion categories manifested on social media platforms serve as the correspondence to real-world societal behaviors and performance. This article presents a novel framework leveraging the Transformer-based pretrained language model for conducting large-scale analysis of publicly available short text data sourced from social media platforms. The research aims to comprehensively understand the dynamic fluctuations across fourteen dimensions of sentiment polarities and emotion categories extracted from short text data expressed by people on public incidents over temporal periods. The study seeks to elucidate the relation between the enactment of relevant policies and the observed sentiment polarities as well as emotion categories. One sentiment polarity and two emotion categories related to policies on public incidents are outlined. This research contributes to the government and pertinent departments by providing insights into the text content on social media platforms concerning public incidents, thereby facilitating the understanding of the evolving sentiment polarities and emotion categories.

I. Introduction

The progressive development of internet technologies has transitioned individuals from passive information consumers to active contributors, enabling everyone to engage in the production and dissemination of media content. Netizens leverage social media platforms as channels to express their assessments, perspectives, and attitudes regarding news, current affairs, and social events [1]. Social media platforms generate tremendous online public opinion data every day, serving as unique conduits for the authentic expression of sentiments and emotions while bridging the chasm between the tangible societal reality and the virtual realm [2]. Through the application of natural language processing methodologies, sentiment classification and emotion recognition of extensive text data sourced from social media platforms can be realized, thereby facilitating the extraction of sentiment polarities and emotion categories prevalent among individuals in response to diverse topics such as public incidents, policies, and consumer products[3]. The temporal evolution of sentiments and emotions expressed by the public on social media platforms maintains significant implications for the government, aiding in the comprehension of public opinions on specific events.