DETEKSI SARKASME DAN IRONI PADA ULASAN LAZADA DI PLAY STORE MENGGUNAKAN MODEL INDOBERT

The advancement of digital technology has transformed consumer behavior, including how users express their opinions through e-commerce platforms such as Lazada. Many reviews are delivered implicitly through sarcasm or irony, making it difficult for conventional sentiment analysis systems to interpret the actual sentiment. This study aims to detect sarcasm and irony in user reviews of the Lazada application on the Google Play Store using the IndoBERT model. A total of 2,000 reviews were collected through web scraping and processed through stages of preprocessing, sentiment classification, and expression labeling. The IndoBERT model was trained to classify three types of expressions: literal, sarcasm, and irony. It achieved an accuracy of 96.40% and an average F1-score of 95.56%. The results show that most reviews were expressed indirectly, with sarcasm being the most dominant expression (995 reviews). These findings indicate that IndoBERT is effective in recognizing complex linguistic nuances and can serve as a reliable tool for sentiment analysis on e-commerce platforms.
Keywords: Sarcasm, Irony, IndoBERT, Sentiment Analysis, Lazada