Sentiment Analysis on Electric Vehicles in Indonesia Using Bidirectional Encoder Representations from Transformers (BERT) and Named Entity Recognition (NER) Methods

Abstract

Air pollution is a major environmental issue due to its significant impact on human health, with the transportation sector being one of the largest contributors. In Indonesia, increasing motor vehicle usage has led to higher greenhouse gas emissions, encouraging the transition toward electric vehicles as a cleaner alternative. However, the adoption of electric vehicles is influenced not only by technical factors such as infrastructure and cost, but also by public perception, which varies across different digital platforms. This study aims to analyze public sentiment toward electric vehicles in Indonesia using a Natural Language Processing (NLP) approach by combining Bidirectional Encoder Representations from Transformers (BERT) and Named Entity Recognition (NER). BERT is utilized to classify sentiments into positive, negative, and neutral categories by considering bidirectional contextual information, while NER is used to identify key entities such as companies, products, locations, and issues discussed in public discourse. The results show that the BERT model achieves an accuracy of 71.05%, precision of 61.31%, recall of 59.28%, and a misclassification error of 28.95%, indicating a fairly good performance in sentiment classification. Furthermore, NER analysis reveals that event and opinion are the most influential factors affecting public interest, followed by company, product, and quality, while location, price, action, and feature have lower influence. Overall, public interest in electric vehicles in Indonesia is relatively high but dynamic, as it is strongly influenced by circulating information and public opinion.