Public opinion analysis of novel coronavirus from online data
Lu Chen, Yang Liu, Yudong Chang, Xinzhi Wang, Xiangfeng Luo
Abstract
Novel coronavirus, now named COVID-19, has swept the world, which is regarded as 'public enemy number one' by WHO. In these months, the coronavirus has become a hot topic and led various public opinion. The traditional strategies for public opinion analyzing seldom take the entities and behaviors into consideration. Focusing on the high fluctuation of public opinion of novel coronavirus event, we propose a Key-Information-oriented Convolutional Neural Network (KIN-CNN) to analyze both relevant entities and behaviors in addition to public opinion trend on Chinese corpus. Firstly, we establish a knowledge set according to the characteristic of distribution in corpus of emotions, behaviors and entities. Secondly, we integrate the other prior knowledge to initialize the convolution kernel for better model performance. Thirdly, as COVID-19 event develops, the dominant public opinion trend is obtained by our approach. Furthermore, the relationship of dominant public opinion with entities and behaviors is established as well in this research.