Research on sentiment classification algorithm for Corona Virus Disease 2019 Weibo text

Abstract

Since the outbreak of Corona Virus Disease 2019, it has had a significant impact on people’s lives. In order to help the government grasp the social opinion and do more scientific and practical propaganda and public opinion guidance for prevention and control, and to fully reflect people’s attitude toward the epidemic and provide data support for government departments to release epidemic prevention measures. This paper uses Corona Virus Disease 2019-related Weibo comments as the research object and analyzes their sentiment using deep learning algorithms. The number of characters in Weibo comments is usually less than 140, which belongs to the category of short texts. Due to the use of few words, random user language, and irregular grammar, these texts have poor performance in text separation and word vector expression, adversely affecting sentiment classification. In order to solve this problem, this paper constructs the BERT-DPCNN model for sentiment analysis of epidemic short texts, which can not only extract the sentence-level text dependencies but also effectively avoid the problem of gradient disappearance of deep neural networks. The experiments show that the BERT-DPCNN model has the best effect and is of great value for the sentiment classification of short epidemic text.