Sentiment Analysis on Reviews of the Documentary Film "Dirty Vote" Using Lexicon-Based and Support Vector Machine Approaches

Abstract

The general election (Pemilu) is a state agenda in Indonesia held every five years. During this democratic event, citizens have the right to freely and fairly choose their leaders. Rules and procedures related to elections are regulated under Law No. 7 of 2017 on General Elections. One of the provisions in this law is the electoral silence period. In the 2024 election, February 11–13, 2024, is designated as the electoral silence period. During this period, Article 287, Paragraph 5 of the Election Law states that print media, online media, social media, and broadcasting institutions are prohibited from disseminating news, advertisements, or any content that benefits or harms election participants. On February 11, 2024, during the silence period, a video titled "Dirty Vote" was uploaded on YouTube, drawing significant public attention. Its release during the silence period sparked controversy and prompted various opinions in the video’s comment section. Sentiment analysis is a suitable method to determine whether public opinions regarding the video are predominantly positive, negative, or neutral. This study utilized the Support Vector Machine (SVM) classification method with different kernels, including linear and non-linear (polynomial, RBF, and sigmoid). To accelerate labeling for large datasets, a Lexicon-Based approach was employed. The combination of SVM and Lexicon-Based methods demonstrated that the linear kernel outperformed others, achieving evaluation metrics of 91.1% accuracy, 91.1% recall, 90.9% precision, and 90.8% F1-score.