A Deep Multi-kernel Uniform Capsule Approach for Hate Speech Detection

ABSTRACT

Hate Speech is an expression that expresses hatred towards people of a specific ethnic group or nationality and incites hatred. Even though many countries have anti-hate speech legislation, hate speech can spread in the native language on social media platforms, resulting in violent riots and protests that spiral out of control and result in anti-social events. Hence, hate speech has caused a crucial social issue. Thus, various intelligent mechanisms have been employed to classify hate speech, depending on the category. A deep learning model has certain limitations for providing n-gram features for text classification of the native language. As a result, in this paper, the Multi-kernel uniform capsule network for multilingual languages is proposed. The proposed method employs a Multi-kernel uniform capsule network to improve feature selection performance by utilizing the capsule network routing algorithm. The experiments were carried out on political, COVID-19 and vaccination, lockdown, and multilingual dataset. The experimental results demonstrate that the proposed methods achieve adequate results when compared with other machine learning models for hate speech detection. © 2022, Springer Nature Switzerland AG.