Sentiment Analysis of COVID-19 on Weibo text using optimized Bi-LSTM model

Abstract The outbreak of Coronavirus disease 2019 (COVID-19) poses a major challenge for China. Sentiment analysis of texts on social media such as Sina-Weibo in China can be useful for adjusting outbreak prevention policies and public health responses. A Bi-LSTM model based on self-attention was used to analyze 110,000 Weibo texts from February 2020 to October 2022, and a topic extraction model for Weibo texts was built using Latent Dirichlet Allocation (LDA). The analysis of the results revealed three key dimensions: time, space, and topic. In terms of time, sentiment among Weibo users was found to be divided into three distinct periods: Formation period, Fluctuation period, and Fading period. In terms of space, the majority of texts about the epidemic on Weibo were concentrated in several key regions, including the Beijing-Tianjin-Hebei region, the Yangtze River Delta, the Pearl River Delta, and the ChengduChongqing urban agglomeration. Sentiment among all the provinces and regions was found to be positive, with Chongqing city and Hunan province having the most positive sentiment. In terms of topic, the main focus of discussions related to the epidemic was found to be ”Impact of COVID-19”, ”Obstacles to work” and ”Life stress”. Consequently, it is suggested that governments adjust their epidemic prevention policies to curb negative sentiment, taking into account changes in sentiment intensity and local concerns. The government is also advised to focus on implementing solutions such as ”Protection against COVID-19”, ”Deregulation”, and ”Well-being benefits for life” by implementing targeted response measures and investing in mental health resources.