{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,8]],"date-time":"2026-06-08T12:04:07Z","timestamp":1780920247101,"version":"3.54.1"},"reference-count":62,"publisher":"Association for Computing Machinery (ACM)","issue":"6","license":[{"start":{"date-parts":[[2026,6,8]],"date-time":"2026-06-08T00:00:00Z","timestamp":1780876800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"funder":[{"DOI":"10.13039\/501100001809","name":"Natural Science Foundation of China","doi-asserted-by":"crossref","award":["U22A2099 and 62336003"],"award-info":[{"award-number":["U22A2099 and 62336003"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Asian Low-Resour. Lang. Inf. Process."],"published-print":{"date-parts":[[2026,6,30]]},"abstract":"<jats:p>Mitigating human-like biases and social stereotypes in pre-trained language models (PLMs) has become a crucial task in Chinese toxic speech detection. While PLMs have achieved state-of-the-art results in mitigating explicit bias (e.g., bias on sensitive attribute words), the study of implicit bias (e.g., bias against specific demographic groups) is still under-explored. Besides, existing debiasing methods focus on the tradeoff between debiasing efficiency and model performance, while ignoring robustness against noisy data. Therefore, a debiasing method that effectively reduces biases while maintaining robustness against noisy data is needed. In this article, we propose Unbiased Contrastive Learning (UCL), which can mitigate explicit and implicit bias while maintaining robustness to noisy data. Specifically, we first analyze the bias representation problem constrained by contrastive learning objective and implement unbiased contrastive objective for learning unbiased text representations to mitigate explicit and implicit biases in Chinese toxic speech detection tasks. UCL inherits the idea of supervised contrastive learning, which encourages representations of the same sensitive attribute to be closer than those of different sensitive attributes and ensures unbiasedness by penalizing the sensitive attribute information contained in the representations. Furthermore, we design conditional normalization to reduce biased classification caused by the imbalanced distribution of demographic groups in the data. Experimental results on Chinese and English datasets show that the proposed method outperforms the state-of-the-art methods and achieves the competitive performance.<\/jats:p>","DOI":"10.1145\/3816029","type":"journal-article","created":{"date-parts":[[2026,5,11]],"date-time":"2026-05-11T11:30:24Z","timestamp":1778499024000},"page":"1-23","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Mitigating Implicit Bias in Chinese Toxic Speech Detection via Unbiased Contrastive Learning"],"prefix":"10.1145","volume":"25","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-1661-6590","authenticated-orcid":false,"given":"Xuan","family":"Feng","sequence":"first","affiliation":[{"name":"Jinan University","place":["Guangzhou, Singapore"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-0512-6730","authenticated-orcid":false,"given":"Yuan","family":"Yao","sequence":"additional","affiliation":[{"name":"Jinan University","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1593-1292","authenticated-orcid":false,"given":"Tianlong","family":"Gu","sequence":"additional","affiliation":[{"name":"Jinan University","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7262-4707","authenticated-orcid":false,"given":"Liang","family":"Chang","sequence":"additional","affiliation":[{"name":"Guangxi Key Laboratory of Trusted Software","place":["Guilin, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6488-9732","authenticated-orcid":false,"given":"Xiaoli","family":"Liu","sequence":"additional","affiliation":[{"name":"Jinan University","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2026,6,8]]},"reference":[{"key":"e_1_3_1_2_2","first-page":"1105","article-title":"Entropy-based attention regularization frees unintended bias mitigation from lists","author":"Attanasio Giuseppe","year":"2022","unstructured":"Giuseppe Attanasio, Debora Nozza, Dirk Hovy, and Elena Baralis. 2022. Entropy-based attention regularization frees unintended bias mitigation from lists. Findings of the Association for Computational Linguistics: ACL 2022 (2022), 1105\u20131119.","journal-title":"Findings of the Association for Computational Linguistics: ACL 2022"},{"issue":"11","key":"e_1_3_1_3_2","first-page":"155","article-title":"Abusive and hate speech classification in Arabic text using pre-trained language models and data augmentation","volume":"23","author":"Badri Nabil","year":"2024","unstructured":"Nabil Badri, Ferihane Kboubi, and Anja Habacha Chaibi. 2024. Abusive and hate speech classification in Arabic text using pre-trained language models and data augmentation. ACM Transactions on Asian and Low-Resource Language Information Processing 23, 11, Article 155 (2024), 1\u201328.","journal-title":"ACM Transactions on Asian and Low-Resource Language Information Processing"},{"key":"e_1_3_1_4_2","doi-asserted-by":"crossref","first-page":"5454","DOI":"10.18653\/v1\/2020.acl-main.485","article-title":"Language (Technology) is power: A critical survey of \u201cBias\u201d in NLP","author":"Blodgett Su Lin","year":"2020","unstructured":"Su Lin Blodgett, Solon Barocas, Hal Daum\u00e9 III, and Hanna Wallach. 2020. Language (Technology) is power: A critical survey of \u201cBias\u201d in NLP. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.5454\u20135476.","journal-title":"In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics."},{"key":"e_1_3_1_5_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W19-3501"},{"key":"e_1_3_1_6_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.aal4230"},{"key":"e_1_3_1_7_2","first-page":"1597","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Chen Ting","year":"2020","unstructured":"Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. In Proceedings of the International Conference on Machine Learning. PMLR, 1597\u20131607."},{"key":"e_1_3_1_8_2","doi-asserted-by":"publisher","DOI":"10.1137\/1.9781611975673.27"},{"key":"e_1_3_1_9_2","doi-asserted-by":"publisher","DOI":"10.1145\/3289600.3291037"},{"key":"e_1_3_1_10_2","first-page":"1","article-title":"FairFil: Contrastive neural debiasing method for pretrained text encoders","author":"Cheng Pengyu","year":"2021","unstructured":"Pengyu Cheng, Weituo Hao, Siyang Yuan, Shijing Si, and Lawrence Carin. 2021. FairFil: Contrastive neural debiasing method for pretrained text encoders. International Conference on Learning Representations (2021), 1\u201312.","journal-title":"International Conference on Learning Representations"},{"key":"e_1_3_1_11_2","first-page":"2736","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2022","author":"Chi Jianfeng","year":"2022","unstructured":"Jianfeng Chi, William Shand, Yaodong Yu, Kai-Wei Chang, Han Zhao, and Yuan Tian. 2022. Conditional supervised contrastive learning for fair text classification. In Findings of the Association for Computational Linguistics: EMNLP 2022. 2736\u20132756."},{"key":"e_1_3_1_12_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.findings-emnlp.58"},{"key":"e_1_3_1_13_2","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00425"},{"key":"e_1_3_1_14_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.emnlp-main.796"},{"key":"e_1_3_1_15_2","doi-asserted-by":"publisher","DOI":"10.1145\/3278721.3278729"},{"key":"e_1_3_1_16_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1045"},{"key":"e_1_3_1_17_2","first-page":"1","volume-title":"Proceedings of the 4th International Conference on Learning Representations","author":"Edwards Harrison","year":"2016","unstructured":"Harrison Edwards and Amos Storkey. 2016. Censoring Representations with an adversary. In Proceedings of the 4th International Conference on Learning Representations. 1\u201314."},{"key":"e_1_3_1_18_2","first-page":"1","article-title":"A minimax approach to supervised learning","volume":"29","author":"Farnia Farzan","year":"2016","unstructured":"Farzan Farnia and David Tse. 2016. A minimax approach to supervised learning. Advances in Neural Information Processing Systems 29 (2016), 1\u20139.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_1_19_2","doi-asserted-by":"crossref","first-page":"9582","DOI":"10.18653\/v1\/2022.emnlp-main.651","volume-title":"Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing","author":"Gaci Yacine","year":"2022","unstructured":"Yacine Gaci, Boualem Benattallah, Fabio Casati, and Khalid Benabdeslem. 2022. Debiasing pretrained text encoders by paying attention to paying attention. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 9582\u20139602."},{"key":"e_1_3_1_20_2","doi-asserted-by":"publisher","DOI":"10.1145\/3580494"},{"key":"e_1_3_1_21_2","doi-asserted-by":"publisher","DOI":"10.1109\/TAFFC.2024.3481419"},{"key":"e_1_3_1_22_2","first-page":"1","volume-title":"International Conference on Learning Representations","author":"Gunel Beliz","year":"2021","unstructured":"Beliz Gunel, Jingfei Du, Alexis Conneau, and Veselin Stoyanov. 2021. Supervised contrastive learning for pre-trained language model fine-tuning. In International Conference on Learning Representations. 1\u201315."},{"key":"e_1_3_1_23_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-long.72"},{"key":"e_1_3_1_24_2","first-page":"6469","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Hayat Munawar","year":"2019","unstructured":"Munawar Hayat, Salman Khan, Syed Waqas Zamir, Jianbing Shen, and Ling Shao. 2019. Gaussian affinity for max-margin class imbalanced learning. In Proceedings of the IEEE\/CVF International Conference on Computer Vision. 6469\u20136479."},{"key":"e_1_3_1_25_2","first-page":"770","volume-title":"Proceedings of the 2024 IEEE Symposium on Security and Privacy (SP)","author":"He Xinlei","year":"2024","unstructured":"Xinlei He, Savvas Zannettou, Yun Shen, and Yang Zhang. 2024. You only prompt once: On the capabilities of prompt learning on large language models to tackle toxic content. In Proceedings of the 2024 IEEE Symposium on Security and Privacy (SP). 770\u2013787."},{"key":"e_1_3_1_26_2","first-page":"1440","volume-title":"Proceedings of the 12th Language Resources and Evaluation Conference","author":"Huang Xiaolei","year":"2020","unstructured":"Xiaolei Huang, Linzi Xing, Franck Dernoncourt, and Michael Paul. 2020. Multilingual Twitter corpus and baselines for evaluating demographic bias in hate speech recognition. In Proceedings of the 12th Language Resources and Evaluation Conference. 1440\u20131448."},{"key":"e_1_3_1_27_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.osnem.2021.100182"},{"key":"e_1_3_1_28_2","doi-asserted-by":"publisher","DOI":"10.1109\/TAFFC.2024.3409830"},{"key":"e_1_3_1_29_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P19-1160"},{"key":"e_1_3_1_30_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N19-1423"},{"key":"e_1_3_1_31_2","first-page":"18661","article-title":"Supervised contrastive learning","volume":"33","author":"Khosla Prannay","year":"2020","unstructured":"Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2020. Supervised contrastive learning. Advances in Neural Information Processing Systems 33 (2020), 18661\u201318673.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_1_32_2","first-page":"3627","volume-title":"Proceedings of the 2024 IEEE 40th International Conference on Data Engineering (ICDE)","author":"Li Jiayang","year":"2024","unstructured":"Jiayang Li, Xuan Feng, Tianlong Gu, and Liang Chang. 2024. Dual-teacher de-biasing distillation framework for multi-domain fake news detection. In Proceedings of the 2024 IEEE 40th International Conference on Data Engineering (ICDE). 3627\u20133639."},{"key":"e_1_3_1_33_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.488"},{"key":"e_1_3_1_34_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P19-1631"},{"key":"e_1_3_1_35_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.coling-main.390"},{"key":"e_1_3_1_36_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.findings-acl.7"},{"key":"e_1_3_1_37_2","doi-asserted-by":"publisher","DOI":"10.1145\/3534678.3539302"},{"key":"e_1_3_1_38_2","unstructured":"Christos Louizos Kevin Swersky Yujia Li Max Welling and Richard Zemel. 2015. The variational fair autoencoder. arXiv:1511.00830. Retrieved from https:\/\/arxiv.org\/abs\/1511.00830"},{"key":"e_1_3_1_39_2","first-page":"16235","volume-title":"Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","author":"Lu Junyu","year":"2023","unstructured":"Junyu Lu, Bo Xu, Xiaokun Zhang, Changrong Min, Liang Yang, and Hongfei Lin. 2023. Facilitating fine-grained detection of Chinese toxic language: Hierarchical taxonomy, resources, and benchmarks. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 16235\u201316250."},{"issue":"3","key":"e_1_3_1_40_2","first-page":"1743","article-title":"The biases of pre-trained language models: An empirical study on prompt-based sentiment analysis and emotion detection","volume":"14","author":"Mao Rui","year":"2022","unstructured":"Rui Mao, Qian Liu, Kai He, Wei Li, and Erik Cambria. 2022. The biases of pre-trained language models: An empirical study on prompt-based sentiment analysis and emotion detection. IEEE Transactions on Affective Computing 14, 3 (2022), 1743\u20131753.","journal-title":"IEEE Transactions on Affective Computing"},{"key":"e_1_3_1_41_2","doi-asserted-by":"publisher","DOI":"10.5555\/3327546.3327583"},{"key":"e_1_3_1_42_2","doi-asserted-by":"publisher","DOI":"10.1145\/2872427.2883062"},{"key":"e_1_3_1_43_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1302"},{"issue":"3","key":"e_1_3_1_44_2","doi-asserted-by":"crossref","first-page":"477","DOI":"10.1007\/s10579-020-09502-8","article-title":"Resources and benchmark corpora for hate speech detection: A systematic review","volume":"55","author":"Poletto Fabio","year":"2021","unstructured":"Fabio Poletto, Valerio Basile, Manuela Sanguinetti, Cristina Bosco, and Viviana Patti. 2021. Resources and benchmark corpora for hate speech detection: A systematic review. Language Resources and Evaluation 55, 3 (2021), 477\u2013523.","journal-title":"Language Resources and Evaluation"},{"key":"e_1_3_1_45_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.647"},{"key":"e_1_3_1_46_2","doi-asserted-by":"crossref","first-page":"16","DOI":"10.1007\/978-3-642-13059-5_5","volume-title":"Proceedings of the Advances in Artificial Intelligence: 23rd Canadian Conference on Artificial Intelligence, Canadian AI 2010, Ottawa, Canada, May 31\u2013June 2, 2010.","author":"Razavi Amir H.","year":"2010","unstructured":"Amir H. Razavi, Diana Inkpen, Sasha Uritsky, and Stan Matwin. 2010. Offensive language detection using multi-level classification. In Proceedings of the Advances in Artificial Intelligence: 23rd Canadian Conference on Artificial Intelligence, Canadian AI 2010, Ottawa, Canada, May 31\u2013June 2, 2010.16\u201327."},{"key":"e_1_3_1_47_2","unstructured":"Akash Saravanan Dhruv Mullick Habibur Rahman and Nidhi Hegde. 2023. FineDeb: A debiasing framework for language models. arXiv:2302.02453. Retrieved from https:\/\/arxiv.org\/abs\/2302.02453"},{"key":"e_1_3_1_48_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W17-1101"},{"key":"e_1_3_1_49_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298682"},{"key":"e_1_3_1_50_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P16-1009"},{"key":"e_1_3_1_51_2","doi-asserted-by":"publisher","DOI":"10.1145\/3711710"},{"key":"e_1_3_1_52_2","unstructured":"Aili Shen Xudong Han Trevor Cohn Timothy Baldwin and Lea Frermann. 2021. Contrastive learning for fair representations. arXiv:2109.10645. Retrieved from https:\/\/arxiv.org\/abs\/2109.10645"},{"key":"e_1_3_1_53_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-long.65"},{"key":"e_1_3_1_54_2","first-page":"1","article-title":"Conditional contrastive learning with kernel","author":"Tsai Yao-Hung Hubert","year":"2022","unstructured":"Yao-Hung Hubert Tsai, Tianqin Li, Martin Q. Ma, Han Zhao, Kun Zhang, Louis-Philippe Morency, and Ruslan Salakhutdinov. 2022. Conditional contrastive learning with kernel. International Conference on Learning Representations (2022), 1\u201317.","journal-title":"International Conference on Learning Representations"},{"key":"e_1_3_1_55_2","first-page":"arXiv\u20132106","article-title":"Conditional contrastive learning: Removing undesirable information in self-supervised representations","author":"Tsai Yao-Hung Hubert","year":"2021","unstructured":"Yao-Hung Hubert Tsai, Martin Q. Ma, Han Zhao, Kun Zhang, Louis-Philippe Morency, and Ruslan Salakhutdinov. 2021. Conditional contrastive learning: Removing undesirable information in self-supervised representations. arXiv e-prints (2021), arXiv\u20132106.","journal-title":"arXiv e-prints"},{"key":"e_1_3_1_56_2","first-page":"683","volume-title":"Proceedings of the International AAAI Conference on Web and Social Media","author":"Vaidya Ameya","year":"2020","unstructured":"Ameya Vaidya, Feng Mai, and Yue Ning. 2020. Empirical analysis of multi-task learning for reducing identity bias in toxic comment detection. In Proceedings of the International AAAI Conference on Web and Social Media. 683\u2013693."},{"issue":"3","key":"e_1_3_1_57_2","first-page":"51","article-title":"Toxic comment classification based on bidirectional gated recurrent unit and convolutional neural network","volume":"21","author":"Wang Zhongguo","year":"2021","unstructured":"Zhongguo Wang and Bao Zhang. 2021. Toxic comment classification based on bidirectional gated recurrent unit and convolutional neural network. ACM Transactions on Asian and Low-Resource Language Information Processing 21, 3, Article 51 (2021), 1\u201312.","journal-title":"ACM Transactions on Asian and Low-Resource Language Information Processing"},{"key":"e_1_3_1_58_2","doi-asserted-by":"publisher","DOI":"10.5555\/2390374.2390377"},{"key":"e_1_3_1_59_2","first-page":"19279","volume-title":"Proceedings of the AAAI Conference on Artificial Intelligence","author":"Wu Yiquan","year":"2024","unstructured":"Yiquan Wu, Yifei Liu, Ziyu Zhao, Weiming Lu, Yating Zhang, Changlong Sun, Fei Wu, and Kun Kuang. 2024. De-biased attention supervision for text classification with causality. In Proceedings of the AAAI Conference on Artificial Intelligence. 19279\u201319287."},{"key":"e_1_3_1_60_2","first-page":"6256","article-title":"Unsupervised data augmentation for consistency training","volume":"33","author":"Xie Qizhe","year":"2020","unstructured":"Qizhe Xie, Zihang Dai, Eduard Hovy, Thang Luong, and Quoc Le. 2020. Unsupervised data augmentation for consistency training. Advances in Neural Information Processing Systems 33 (2020), 6256\u20136268.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_1_61_2","first-page":"325","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Zemel Rich","year":"2013","unstructured":"Rich Zemel, Yu Wu, Kevin Swersky, Toni Pitassi, and Cynthia Dwork. 2013. Learning fair representations. In Proceedings of the International Conference on Machine Learning. 325\u2013333."},{"key":"e_1_3_1_62_2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.232"},{"key":"e_1_3_1_63_2","first-page":"1","volume-title":"Proceedings of the 2021 IEEE International Conference on Multimedia and Expo (ICME)","author":"Zhu He","year":"2021","unstructured":"He Zhu and Shan Yu. 2021. Intra-class uncertainty loss function for classification. In Proceedings of the 2021 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 1\u20136."}],"container-title":["ACM Transactions on Asian and Low-Resource Language Information Processing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3816029","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,8]],"date-time":"2026-06-08T11:25:26Z","timestamp":1780917926000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3816029"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,8]]},"references-count":62,"journal-issue":{"issue":"6","published-print":{"date-parts":[[2026,6,30]]}},"alternative-id":["10.1145\/3816029"],"URL":"https:\/\/doi.org\/10.1145\/3816029","relation":{},"ISSN":["2375-4699","2375-4702"],"issn-type":[{"value":"2375-4699","type":"print"},{"value":"2375-4702","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,8]]},"assertion":[{"value":"2025-02-05","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-05-02","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-06-08","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}