The Role of Context in Detecting the Target of Hate Speech (ACL Anthology)

With an emphasis on Dutch-language social media, the current study tackles the ongoing problem of contextual sensitivity in automated hate speech identification. It specifically looks into whether nasty remarks on Facebook are directed at migrants or other groups. The authors show that contextual integration significantly improves classification accuracy by manually annotating conversational context and integrating it into the Dutch transformer-based model BERTje. The results highlight how crucial complex contextual encoding is to enhancing the accuracy of hate speech detection systems.

https://aclanthology.org/2022.trac-1.5