Research on sentiment analysis and entity recognition of COVID-19 based on multi-task sentiment analysis model in artificial intelligence

Zhang, X.; Xu, Z.

Text sentiment analysis is an important branch of artificial intelligence. After the outbreak of COVID-19, a lot of highly emotional texts were flooded on the Internet. In the task of entity-level sentiment analysis for major public health events, there is often a lack of large-scale labeled corpus and a mature model that can effectively deal with medical and health texts. A multi-task sentiment analysis model based on BERT was proposed, which combined sentiment analysis and entity recognition. This model is used to extract sentiment information of Weibo comments related to COVID-19 and judge whether the entities in the text are targeted by emotions. The composition layer was constructed by combining the emotional information and entity information, and the loss function was designed to integrate the multi-task into the same model. More than 5000 Weibo text data about COVID-19 were preprocessed and annotated to form an entity-level corpus. Finally, several common methods were used to compare with the proposed method. Several comparison indexes were used to evaluate the model. The experimental results showed that the proposed model had better overall performance than other models. © 2021 IEEE.