Zaenal Abidin, Nala Adina, Riza Arifudin, Aji Purwinarko, Hamdani Hamdani, David Leandro Wibisono
The exponential growth of social media users has changed the way people express their thoughts online. The freedom of expression offered by social media raises several issues. A major issue is the increasing number of hate speech posts containing offensive and foul language. Hate speech posts are targeted to individuals or groups of communities or organizations. This paper presents the use of a deep learning method to conduct multi-label text classification for hate speech tweets detection, including detecting the target, category, and degree of hate speech in the Indonesian language. The Convolutional Neural Network (CNN) method was employed in this current study. This study also implemented Word2Vec as word embedding. The result shows that the implementation of Word2Vec improves detection accuracy by 7.12%. The accuracies of using CNN and the CNN+Word2Vec are 64.07% and 71.19%, respectively. © 2023 Author(s).
Department of Computer Science, Universitas Negeri Semarang, Semarang, Indonesia; Department of Informatics, Mulawarman University, Samarinda, Indonesia
