Using Twitter Data for Telecommunication Service Improvement: A Case Study of Telkomsel Indonesia

Abstract

Recently, many companies and governments have made Twitter accounts, providing an outlet for the users and citizens to voice their opinions. With the large number of tweets such an account receives, a system needs to be established to automatically analyze them. While Indonesia has over 18 million Twitter users, there exists little research on sentiment analysis and geocoding of Indonesian tweets. This paper proposes an approach to map negative sentiment tweets, allowing companies and governments to locate specific areas requiring their attention. The sentiment analysis is performed with the IndoBERTweet model, while the location extraction is done, based on both geotags and geocoding. It is demonstrated via experiments that the proposed system could successfully map tweets with negative sentiments, using the extracted locations.