Sentiment Analysis of Covid-19 As A Social Media Pandemic

Abstract

A large amount of information about Covid-19 that spreads quickly can lead to a perception of opinion and sentiment for those who read it. This research studies how text networking is formed, sentiment analysis and topics modelling that is widely discussed related to the Covid-19 theme. The text networking analysis was carried out on data taken from 4 different times, namely on 26 March, 29 March, 28 June and 23 July 2020 giving the result that the largest edge, nodes and modularity were in the conversation data on July 23, 2020. Sentiment analysis shows how the public responds to the Covid-19 pandemic. Sentiment analysis from tweet data in March 2020 showed 51% as positive sentiment and 49% as negative sentiment, with an accuracy rate of 0.7586, specificity 0.6667, prevalence 0.5862. Then tweet data in June 2020 showed 59% as negative sentiment and 41% as positive sentiment, with an accuracy rate of 0.6486, specificity 0.6111, prevalence 0.5135. Analysis of topic modelling has succeeded in collecting words related to certain topics, such as the data on March 26, 2020, representing talks related to the topic of "doing activities from home", "health", and "government policy". The data on March 29, 2020, represent talks related to the topic of "activities from home", "expression of feelings", "new habits". The data on June 28, 2020, represent talks related to the topic of "health protocol", "social assistance", "health". And on July 23, 2020 data represents talks related to the topic of "data security", "fine policy", and "policy".