Sentiment Analysis of Facebook Posts through Special Reactions: The Case of Learning from Home in Indonesia Amid COVID-19

Abstract

In contrast to several other countries, Indonesian sentiment analysis research is primarily focused on the text-based analysis of Twitter. Given that Twitter users in Indonesia account for less than a seventh of those on Facebook, sentiment analysis on the latter may have a greater impact than on the former. This research sought to close that gap in the literature by pioneering the use of Facebook special reactions as an alternative to text-based sentiment analysis on social media posts about Indonesian social issues. The topic of learning from home in the midst of the COVID-19 pandemic was chosen because it is both timely and relatable to almost everyone in the country. Through CrowdTangle, a total of 39,657 Facebook posts containing the key phrase “belajar dari rumah” were gathered, but only 9,310 of them received special reactions and thus remained to be analyzed quantitatively. The results indicated that with the exception of ‘love,’ all special reactions are somewhat correlated, suggesting that they can be used to indicate the negative valence of a Facebook post. Further analysis revealed a significant increase in the proportion of posts with a negative valence during the second year of the COVID-19 pandemic. The textual analysis of the posts revealed that those with a negative valence primarily discuss internet access and other IT infrastructure issues that presumably impede learning from home activities for some. The main contribution of this study is to demonstrate how to analyze special reactions on Facebook for sentiment analysis purposes, particularly in the context of Indonesia. Additionally, it lays out how Facebook's special reactions have the potential to be used in conjunction with text-based sentiment analysis to provide a complete picture of the social issue being investigated.