TY - GEN
T1 - Hate Speech and Counter Speech Detection
T2 - 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2022
AU - Yu, Xinchen
AU - Blanco, Eduardo
AU - Hong, Lingzi
N1 - Funding Information:
This work was supported by Research Seed Grant of the UNT College of Information. We would like to thank the anonymous reviewers for their insightful comments and suggestions.
Publisher Copyright:
© 2022 Association for Computational Linguistics.
PY - 2022
Y1 - 2022
N2 - Hate speech is plaguing the cyberspace along with user-generated content. This paper investigates the role of conversational context in the annotation and detection of online hate and counter speech, where context is defined as the preceding comment in a conversation thread. We created a context-aware dataset for a 3-way classification task on Reddit comments: hate speech, counter speech, or neutral. Our analyses indicate that context is critical to identify hate and counter speech: human judgments change for most comments depending on whether we show annotators the context. A linguistic analysis draws insights into the language people use to express hate and counter speech. Experimental results show that neural networks obtain significantly better results if context is taken into account. We also present qualitative error analyses shedding light into (a) when and why context is beneficial and (b) the remaining errors made by our best model when context is taken into account.
AB - Hate speech is plaguing the cyberspace along with user-generated content. This paper investigates the role of conversational context in the annotation and detection of online hate and counter speech, where context is defined as the preceding comment in a conversation thread. We created a context-aware dataset for a 3-way classification task on Reddit comments: hate speech, counter speech, or neutral. Our analyses indicate that context is critical to identify hate and counter speech: human judgments change for most comments depending on whether we show annotators the context. A linguistic analysis draws insights into the language people use to express hate and counter speech. Experimental results show that neural networks obtain significantly better results if context is taken into account. We also present qualitative error analyses shedding light into (a) when and why context is beneficial and (b) the remaining errors made by our best model when context is taken into account.
UR - http://www.scopus.com/inward/record.url?scp=85138357990&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85138357990&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85138357990
T3 - NAACL 2022 - 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference
SP - 5918
EP - 5930
BT - NAACL 2022 - 2022 Conference of the North American Chapter of the Association for Computational Linguistics
PB - Association for Computational Linguistics (ACL)
Y2 - 10 July 2022 through 15 July 2022
ER -