PERSONA: Personality-based deep learning for detecting hate speech
Kyuhan Lee, Sudha Ram · Journal of the Association for Information Systems · 2020
Hate speech in an online environment has detrimental impacts on the wellbeing of individuals, online communities, and social network platforms. Consequently, the automated detection of hate speech have become a significant issue for various stakeholders. While previous studies have proposed many approaches for this issue, we find an important research gap that they have neglected a plethora of studies from psychology investigating the relationship between personality and hate. To fill the gap, we adopt a text-mining approach which fully automates the process of personality inference. Based its results, we build a personality-based deep learning model for detecting online hate speech (i.e., PERSONA). We validated our model with two real-world cases. The results show that our model significantly outperforms state-of-the-art baselines including a method proposed by Google. Our study paves the way for future research by incorporating psychological aspects into the design of a deep-learning model for hate speech detection.