Sentiment Analysis of Community Response Indonesia Against Covid-19 on Twitter Based on Negation Handling

Abstract

The use of the internet globally, especially on the use of social media, includes Indonesia as one of the most active users in the world. The amount of information that can be obtained can be used to be processed into useful information, for example, information about the public sentiment on a particular topic. Tracking and analyzing tweets can be a method to find out people's thoughts, behavior, and reactions regarding the impact of Covid-19. The key to sentiment analysis is the determination of polarity, which determines whether the sentiment is positive or negative. The word negation in a sentence can change the polarity of the sentence so that if it is not handled properly it will affect the performance of the sentiment classification. In this study, the implementation of negation handling on sentiment analysis of Indonesian people's opinions regarding COVID-19 on Twitter has proven to be good enough to improve the performance of the classifier. Accuracy results obtained are 59.6% compared to adding negation handling accuracy obtained is 59.1%. Although the percentage result is not high, documents that include negative sentences have more meaning than negative sentences. However, for the evaluation using the MCC evaluation matrix, the results were quite good for the testing data. For the results of the proposed method whether it is suitable for data that has two classes or three classes when viewed from the results of the evaluation matrix, the proposed method is more suitable for binary data or data that has only two classes.