Sentiment Analysis Model for KlikIndomaret Android App During Pandemic Using Vader and Transformers NLTK Library

Abstract

COVID-19 has changed the Indonesian people's shopping habits for consumer goods. The online retail application came as a response to social distancing and stay-at-home advice. KlikIndomaret is an online retail application that uses the omnichannel concept. As the number of downloads increased, the number of various comments and sentiments on that application also increased. In this study, the researcher did a sentiment analysis aimed to improve the quality of application experiences and retail services. The result of the analysis reflected the services given to customers thus far. The data included reviews and star ratings derived from 4,066 reviews which went under the process of data pre-processing. The methods used in this study were VADER and NLTK, improved by Transformer, without pre-training data. These methods could filter the users' reviews with sarcasm tone. The results were sentiment labels that were appropriate based on the score comparison of positive and negative sentiments in one user's review. This approach made the review sentiment process of thousands of data faster and more accurate.