Annotation System to Build Cyberbullying and Hate Speech Detection Model Training Dataset

Abstract

During 2019, Indonesian people experienced the election period which triggers many hate speech and cyberbullying cases on Twitter. A detection tool to screen social media data can be used to avoid the spread of negative content. A supervised machine learning approach can be used to build this detection tool. However, it needs thousands of labeled data to develop the machine learning model with high accuracy. In the current study phase, an annotation system was proposed to help the researchers to label Twitter raw data collected in the previous phase. An open-source tool was utilized to build a user-friendly web-based system. Three main features are proposed including multi labels annotation, multi-users validation, and dashboard page. This system can help the annotators to perform labeling task for thousands of text data.