Analisis Sentimen Ujaran Kebencian pada Instagram Tokoh Politik Indonesia Menggunakan Metode Bert

Abstract

The policies made by Indonesian political figures elicited various responses from the public, one of which was hate speech. Hate speech is a serious problem in Indonesia that can easily be raised on any social media, one of which is the Instagram social media comment column. This research uses sentiment analysis to detect whether there is hate speech in the comment column of an Indonesian political figure. The research stage of this research was to collect data in Anies Baswedan's Instagram comment column using crawling techniques that were collected into datasets, then pre-processing the data to clean the dataset. Furthermore, the cleaned data will be translated into English and labeled as positive, neutral and negative using the BERT method, namely bert-base-multilingual-uncased. After the labeling process is complete, a negative dataset will be taken which will be classified again into hate speech and not hate speech using indobert-base-p2. Research using indobert-base-p2 resulted in an accuracy of 90% for 3 epochs and 84% for 10 epochs.