An optimization of a Lexicon Based Sentiment Analysis Method on Indonesian App Review

Over the past years, the popularity of mobile applications (or known as mobile apps) is continuously growing from year to year in the last decade. The development of mobile apps now becomes a part of business activity. It is essential for each developers to understand the users’ need to keep their users on using their app. Sentiment analysis or opinion mining is an approach that can be used to analyze and conclude the opinion of the users of mobile apps. Recent method using lexicon based approach which is proposed by the previous study still has poor performance and still can be improved. There are two optimization chances that can be explored: lexicon resource usage evaluation and the application of domain specific features. This study tries to explore these two optimization chances in order to improve and optimize the method proposed by the previous study. The result shows that SentiWordNet can outperform other lexicon resources and the additional of domain specific features has a good impact on the performance of the classifier both on overall accuracy and f-measure average evaluation parameters compared to not only the results of this study but also compared to the result of the previous study.