Enhancing Hate Speech Detection in Social Media Using IndoBERT Model: A Study of Sentiment Analysis during the 2024 Indonesia Presidential Election
Ramadhan Ihsani Yulfa, Benediktus Hengki Setiawan, Gerry Gilbert Lourensius, Kartika Purwandari · 2023
Legislative elections, presidential elections, and village head elections are all on the schedule for Indonesia in 2024. Law No. 7 of 2017 governs these elections, emphasizing the ideals of direct, public, free, confidential, honest, and fair elections. With the growth of social media, political personalities are employing these channels for campaign activity, resulting in a variety of user responses, including hate speech. Detecting hate speech is critical for preventing and mitigating the bad impact it may have on society. The purpose of this article is to investigate the usage of IndoBERT, a fine-tuned BERT model, to improve hate-speech identification during the 2024 Indonesia Presidential Election. The research focuses on hate speech in the context of the presidential election and the challenges raised by social media. In this paper, we utilized Twitter as a case study to develop effective hate-speech identification techniques. The method comprises fine-tuning IndoBERT using a dataset of Indonesian tweets on the election that have been preprocessed to reduce noise and classified as hate speech or non-hate speech using a pre-trained model. The accuracy, precision, recall, and F1-score of the model are used to assess its performance. The results reveal that the IndoBERT model detects hate speech in Indonesian Twitter data with high accuracy. A future study might look towards fixing typos in the dataset and examining expert recognition of hate speech scenarios.