BACKGROUND
From the approval of the first COVID-19 vaccine to the end of the COVID-19 pandemic in China, COVID-19 vaccines are considered one of the most effective ways to protect people from the COVID-19 pandemic. The public has been using social media platform such as Sina Weibo as an outlet to express their feelings, concerns, and opinions about COVID-19 vaccines. This study examined COVID-19 vaccine-related blogs between the Dec 2020 and Oct 2021, when large-vaccination were carried out.
OBJECTIVE
The aim of this study was to examine key themes and topics from COVID-19 vaccine blogs in China, as well as to explore the trends and to detect the peaks and events in public opinions and sentiments during the vaccination.
METHODS
A corpus of 893,923 blogs were captured on the largest social platform Sina Weibo in China between Dec 2020 and Oct 2021. There were several blogs classification tasks: first, they need to be classified as relevant or irrelevant; then the relevant blogs were classified as objective and subjective; at last, the subjective blogs were classified as positive, negative, or neutral. We manually annotated 1800 blogs in each classification tasks, respectively. The annotated blogs were used to train a BERT-based classifier. Topic modelling was applied on the total valid blogs and the sentiment blogs to identify the trends and attitudes of the public towards vaccination. Correlations between blogs or sentiments and daily vaccination cases were calculated. We also identified the sentiment peak and the critical public events that may have triggered the surges.
RESULTS
The accuracy for classification task of relevant & irrelevant, subject & object, emotional polarity are 97.2%, 96.3%, and 85.3%, respectively. By summarizing the topics chronically, it is observed that the vaccine-related discussion topics changed with relation to the critical events about vaccine. The public was more interested in the vaccination process than in the emotion to the vaccine. The topics about the negative sentiment were mainly about the side effect and vaccination hesitancy.
CONCLUSIONS
This paper applied Bert-based classifier and natural language processing methods to examine the changes of the topics and emotions about the COVID-19 vaccines in China during the pandemic. To identify the trends of the topics and sentiments could provide the deep insights for policymakers and health officials to construct appropriate policies and programs for COVID-19 vaccination and future pandemics.