RUU Perampasan Aset merupakan isu yang banyak diperbincangkan masyarakat di media sosial TikTok. Karakteristik komentar yang dapat memuat lebih dari satu aspek menyebabkan analisis sentimen secara umum belum mampu menggambarkan opini publik secara rinci. Penelitian ini bertujuan menganalisis arah opini publik terhadap isu RUU Perampasan Aset menggunakan metode Lexicon-Based Sentiment Analysis dan Multi Aspect-Based Sentiment Analysis (M-ABSA). Data penelitian berupa 2.971 komentar TikTok yang diperoleh melalui web scraping menggunakan Apify. Proses analisis meliputi preprocessing, identifikasi aspek menggunakan POS Tagging dan Trigger Word, ekstraksi kata opini menggunakan context window, serta penentuan sentimen menggunakan kamus lexicon. Evaluasi dilakukan untuk membandingkan tiga skenario identifikasi aspek dengan menggunakan nilai Micro F1-score sebagai acuan utama. Hasil evaluasi menunjukkan bahwa Skenario 1 berbasis POS Tagging memperoleh rata-rata Micro F1-score sebesar 14,68%, Skenario 2 berbasis Trigger Word sebesar 67,27%, dan Skenario 3 berbasis Hybrid (POS Tagging Trigger Word) sebesar 70,74%. Berdasarkan hasil tersebut, Skenario 3 Hybrid yang mengombinasikan POS Tagging dan Trigger Word dipilih sebagai skenario terbaik. Hasil analisis menunjukkan bahwa opini publik terhadap isu RUU Perampasan Aset cenderung negatif. Hasil tersebut menunjukkan bahwa kombinasi Lexicon-Based Sentiment Analysis dan M-ABSA mampu menganalisis sentimen pada setiap aspek dalam komentar TikTok terkait isu RUU Perampasan Aset.


Deskripsi Lain

RUU Perampasan Aset is an issue that has been widely discussed by the public on TikTok social media. The characteristics of comments, which may contain more than one aspect, cause general sentiment analysis to be unable to describe public opinion in detail. This study aims to analyze the direction of public opinion on the RUU Perampasan Aset issue using the Lexicon-Based Sentiment Analysis and Multi Aspect-Based Sentiment Analysis (M-ABSA) methods. The research data consist of 2,971 TikTok comments obtained through web scraping using Apify. The analysis process includes preprocessing, aspect identification using POS Tagging and Trigger Word, opinion word extraction using a context window, and sentiment determination using a lexicon dictionary. Evaluation was conducted to compare three aspect identification scenarios using the Micro F1-score the main evaluation metric. The evaluation results show that Scenario 1 based on POS Tagging achieved an average Micro F1-score of 14.68%, Scenario 2 based on Trigger Word achieved 67.27%, and Scenario 3 based on the Hybrid approach (POS Tagging Trigger Word) achieved 70.74%. Based on these results, Scenario 3 Hybrid, which combines POS Tagging and Trigger Word, was selected the best scenario. The analysis results indicate that public opinion on the RUU Perampasan Aset issue tends to be negative. These findings show that the combination of Lexicon-Based Sentiment Analysis and M-ABSA is capable of analyzing sentiment for each aspect in TikTok comments related to the RUU Perampasan Aset issue.