Klasifikasi Komentar Rasis Menggunakan Bidirectional Long Short-Term Memory - Convolutional Neural Network Berbasis Ekstensi Web Browser

Abstract

Racism remains a serious social issue that emerges from cultural diversity in multicultural countries such as Indonesia. Acts of racism can appear in various forms, including racist expressions conveyed online. Social media platforms like Instagram Comment Section often become vulnerable spaces for such behavior. This issue is further exacerbated by Instagram’s moderation system, which is not yet sufficiently effective in filtering racist comments. This research aims to develop a system that classifies racist comments on Instagram by implementing a Natural Language Processing (NLP)-based model using BiLSTM-CNN algotithm within a Web Browser Extension that operates on Instagram’s comment page. By integrating the browser extension with a model as backend, the system can efficiently classify short-text comments commonly found on Instagram. The use of a browser extension also serves as a practical solution due to Instagram’s lack of publicly accessible APIs, enabling the manipulation of webpage elements (HTML) to apply the classification results directly on the platform frontend. The model was evaluated using a Confusion Matrix and through three data-splitting scenarios: 70%-30%, 80%-20%, and 90%-10%. The 90%-10% split achieved the highest performance with an accuracy of 0.91, outperforming the other splits (0.89 for 80%-20% and 0.88 for 70%-30%). With its high accuracy, strong evaluation results, and proven capability to classify comments via the browser extension on Instagram’s web platform, this system is expected to serve as an effective alternative tool for helping users filter racist comments and fostering a healthier social media environment.