Investigation of challenges in aspect-based sentiment analysis enhanced using softmax function on twitter during the 2024 Indonesian presidential election

• Published in: Procedia Computer Science
• Main Author: Tanoto K.; Gunawan A.A.S.; Suhartono D.; Mursitama T.N.; Rahayu A.; Ariff M.I.M.
• Format: Conference paper
• Language: English
• Published: Elsevier B.V. 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85213053252&doi=10.1016%2fj.procs.2024.10.327&partnerID=40&md5=e27a591caff057943a369e3d1272b82a

In recent years, social media platforms have emerged as powerful channels for political discourse and public opinion expression. Twitter has become a dynamic space where individuals share views, opinions, and sentiments on various topics, including significant political events such as elections. The 2024 Presidential Election in Indonesia is a good moment to represents public sentiment. The study aims to the explore sentiment expression on social media, considering diverse perspectives, language nuances, and specific cultural contexts within the Indonesian political landscape. To determine sentence sentiment, both standard sentiment analysis and ABSA are considered, with a preference for ABSA due to its ability to identify aspects within a sentence and provide more detailed sentiment analysis. Therefore, this research proposes a solution by applying a BERT-based deep learning model. The model's performance is evaluated using standard metrics, including confusion matrix, precision, recall, F1-score, and accuracy. Comparative assessments are made with machine learning models, such as Random Forest, Naïve Bayes, and Support Vector Machine. The results indicate that, with a 50:50 aspect ratio, BERT, Random Forest, Naïve Bayes, and Support Vector Machine achieve comparable accuracies. However, when the aspect ratio shifts to 60:40, BERT outperforms other models. Notably, in scenarios with a larger amount of training data, Support Vector Machine demonstrates slightly superior performance in predicting sentiment from Twitter data. © 2024 The Authors.