Why Is It Hate Speech? Masked Rationale Prediction for Explainable Hate Speech Detection

Proceedings of the 29th International Conference on Computational Linguistics, pages 6644–6655 October 12–17, 2022. 6644 Why Is It Hate Speech? Masked Rationale Prediction for Explainable Hate Speech Detection Jiyun Kim, Byounghan Lee, Kyung-Ah Sohn∗ Ajou University {hamjee66, qudgks96, kasohn}@ajou.ac.kr ∗Corresponding author Abstract In a hate speech detection model, we should consider two critical aspects in addition to de- tection performance–bias and explainability. Hate speech cannot be identified based solely on the presence of specific words; the model should be able to reason like humans and be explainable. To improve the performance con- cerning the two aspects, we propose Masked Rationale Prediction (MRP) as an intermedi- ate task. MRP is a task to predict the masked human rationales–snippets of a sentence that are grounds for human judgment–by referring to surrounding tokens combined with their un- masked rationales. As the model learns its reasoning ability based on rationales by MRP, it performs hate speech detection robustly in terms of bias and explainability. The proposed method generally achieves state-of-the-art per- formance in various metrics, demonstrating its effectiveness for hate speech detection. Warn- ing: This paper contains samples that may be upsetting. 1 Introduction With the recent development of social media and online communities, hate speech, one of the criti- cal social problems, can spread easily. The spread of hate strengthens discrimination and prejudice against the target social groups and can violate their human rights. Moreover, online hatred extends of- fline and causes real-world crimes. Therefore, prop- erly regulating online hate speech is important to address many social problems related to aversion. In addition to the detection performance, two es- sential considerations are involved in implementing a hate speech detection model–bias and explain- ability. Hate speech should not be judged by any specific word but by the context in which the word is used. Even if any word generally considered vicious does not exist in a text, the text can be hate speech. A specific expression does not always im- ply hatred either (e.g., e.g., ‘nigger’) (Del Vigna12 Figure 1: Examples for the two methods to get the final ground truths. Example input sentences are represented with the class and human rationale labels. In this fig- ure, HateXplain uses the same ground truth about both normal and hateful sentences for the loss. However, our method could determine the two classes with the ground truths. et al., 2017). However, the presence of this word can cause a model to make a biased detection of hate speech. This erroneous judgment may inadver- tently strengthen the discrimination against the tar- get group of the expression (Sap et al., 2019; David- son et al., 2019). In this respect, the model’s bias toward specific expressions should be excluded. The expressions that can cause biased judgment should be interpreted in context. It means it is vital for the hate speech detection models to have the ability to make judgments based on context, as humans do. Therefore, the model should be explainable to humans so that the rationale behind a result …

Content truncated to 500 words