JMIR Publications

ABSTRACT

Background:

The December 2019 outbreak of a novel coronavirus (COVID-19) in Hubei Province, China, resulted in strong behavioral restrictions in other countries through lockdowns, with adverse psychological effects on the identified public. However, since the spread of COVID-19, sentiment analysis, which analyzes the feelings of individuals based on text information, has been used to reflect the psychological changes in individuals and public concerns as well as to provide information for public health and policy. While infodemiology is being actively promoted and advanced, few studies related to social media have been conducted in Japan, although they are used by people of all ages.

Objective:

In this study, infodemiology sentiment analysis is used to understand popular sentiment toward COVID-19 on social media, particularly in Japan.

Methods:

We conducted a day-by-day analysis of Twitter data using 4,894,009 tweets containing the keywords "corona," "COVID-19," and "new pneumonia" from March 23 to April 21, 2020, approximately two weeks before and after the first declaration of a state of emergency in Japan. We also processed tweet data into vectors for each word, employed the Fuzzy-C-Means (FCM) method, a type of cluster analysis, for the words in the sentiment dictionary, setting up seven sentiment clusters (negative: anger, sadness, surprise, disgust; neutral: anxiety; positive: trust and joy), and conducted sentiment analysis divided into tweet groups and retweet groups.

Results:

According to sentiment analysis, positive feelings of "joy" and negative feelings such as "worry" and "disgust," were prevalent in the early stages of COVID-19 spread. In addition, a comparison of the ratio of positive and negative emotions between self-published tweets and retweets (RTs), which were mainly used to spread information, showed that RTs tended to post more negative emotions.

Conclusions:

The results suggest that expectations about infection prevention measures owing to the declaration of a state of emergency in the first round led to an increase in positive values, while topics related to COVID-19 tended to spread information with more negative content. Based on the results of this study, it is necessary to conduct a network analysis of the topics included in each emotion based on the results of the sentiment analysis, as well as a more in-depth analysis in infodemiology.