[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-137534-105":59,"doc-detail-137534-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","cracking-the-code-enhancing-implicit-hate-speech-detection-through-coding-classification","Cracking the Code - Enhancing Implicit Hate Speech Detection through Coding Classification","","The internet has become a hotspot for hate speech (HS), threatening societal harmony and individual well-being. Automatic detection excels at explicit hate speech, yet struggles with subtler implicit hate speech (im-HS). This work introduces a new taxonomy for im-HS detection, defining six encoding strategies, codetypes. Two methods integrate codetypes into im-HS detection: direct LLM prompting for classification and using LLMs as encoders with codetypes embedded. Experiments show codetypes improve im-HS detection on Chinese and English datasets, validating cross-lingual effectiveness.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/cracking-the-code-enhancing-implicit-hate-speech-detection-through-coding-classification/137534/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/cracking-the-code-enhancing-implicit-hate-speech-detection-through-coding-classification/137534.png","ImageObject",300,407,{"name":92,"@type":93},"Quinn Holloway","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-04","2026-08-22",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",15,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does the paper address?","Question",{"text":113,"@type":114},"The paper targets implicit hate speech (im-HS) detection, which is difficult for models compared with explicit hate speech due to subtler encoding strategies.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What are codetypes in this work?",{"text":118,"@type":114},"Codetypes are six encoding strategies defined in the paper as rhetorical tactics extracted from im-HS, used to enhance detection.",{"name":120,"@type":111,"acceptedAnswer":121},"How are codetypes integrated into detection systems?",{"text":122,"@type":114},"The paper presents two approaches: prompting large language models to classify sentences from generated responses, and using LLMs as encoders with codetypes embedded during encoding.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},137534,1787416826,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":105,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":130,"read_time":144,"preview_page_count":4},2336474466712,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","Cracking the Code: Enhancing Implicit Hate Speech Detection through Coding Classification  \nLu Wei 1 , Liangzhi Li 1 * , Tong Xiang 1 , Xiao Liu2 , Noa Garcia 1  \n1The University of Osaka, Osaka, Japan  \n2Meetyou AI Lab, Xiamen, China  \n{lu-wei, [tongxiang}@is.ids.osaka-u.ac.jp](tongxiang}@is.ids.osaka-u.ac.jp)  \n{li, [noagarcia}@ids.osaka-u.ac.jp](noagarcia}@ids.osaka-u.ac.jp), [runnishino@gmail.com](runnishino@gmail.com)  \nAbstract  \nThe internet has become a hotspot for hate speech (HS), threatening societal harmony and individual well-being. While automatic detection methods perform well in identifying explicit hate speech (ex-HS), they struggle with more subtle forms, such as implicit hate speech (im-HS) . We tackle this problem by introducing a new taxonomy for im-HS detection, defining six encoding strategies named codetypes.  \nWe present two methods for integrating codetypes into im-HS detection: 1) prompting large language models (LLMs) directly to classify sentences based on generated responses, and  \n2) using LLMs as encoders with codetypes embedded during the encoding process. Experiments show that the use of codetypes improvesim-HS detection in both Chinese and English datasets, validating the effectiveness of our approach across different languages.  \nNOTE: The samples presented in this paper may be considered offensive or vulgar.  \n1 Introduction  \nIn the current socio-cultural context, the identification of hate speech (HS) has become increasingly important (Das et al., 2020 ; Weidinger et al., 2022 ; Yin and Zubiaga, 2022) . Numerous studies highlight the negative impact of toxic language and HS, not only on the directly targeted individuals (Jikeliet al., 2023 ; Hettiachchi et al., 2023 ; Miller Yoder et al., 2023 ; Klutse et al., 2023 ; Sharma et al., 2022 ; Xu and Weiss, 2022) but also on the society as a whole (Erjavec and Kovai, 2012 ; Saha et al., 2019 ; Kiritchenko et al., 2021 ; Rapp, 2021 ; Maarouf et al., 2022 ; Aleksandric et al., 2022) . HS has the potential to exacerbate divisions and conflicts (Schmitz et al., 2022), and in extreme cases, can threaten community stability (Perez and Karmakar, 2023 ; Williams et al., 2020) .  \n* Corresponding author.  \nAbbreviation  \ntxl是社会问题，应该有人管，而不是当透明人  \nFigure 1: Selected examples from the ToxiCN dataset (Lu et al., 2023) that illustrate six codetypes. English translation below is provided for clarity. The keywords that encode specific types of implicit hate information are highlighted.  \nWithin the prevailing research, the majority of scholarly efforts are dedicated to categorizing macro-level concepts of HS (Jiang et al., 2023 ; Choi et al., 2023 ; Sarwar and Murdock, 2022 ; Alexander and Wang, 2023), which can generally be classified into two types: explicit hate speech (ex-HS) (Schmidt and Wiegand, 2017) and implicit hate speech (im-HS) (ElSheriefet al., 2021) . ex-HS refers to straightforward toxic statements, typically featuring derogatory language (Gao et al., 2017 ; Waseem and Hovy, 2016) . In contrast, im-HS does not contain direct expressions of hate, being a more subtle form to convey prejudice, discrimination, or hatred towards a specific group through sarcasm, insinuation, or other obscured means (ElSherief et al., 2021 ; Wright et al., 2021 ; Huang et al., 2023b) .  \nWith the increasing spread of HS on the internet, online platforms have started to control its dissemination (Twitter, 2023), a focus area within content moderation of social bots (Venkatesh et al., 2024 ; Park et al., 2024) . Due to the explicit na-  \n112  \nProceedings of the 5th Workshop on Trustworthy NLP (TrustNLP 2025) , pages 112–126 May 3, 2025 ©2025 Association for Computational Linguistics  \nture of ex-HS, detection methods can achieve high detection rates (Lu et al., 2023 ; Roychowdhury and Gupta, 2023 ; Caselli et al., 2020) . In contrast, im-HS involves sophisticated encoding rules that make it easier to evade automatic detection (Gunturi et al., 2023 ; Wiegand et al., 20","cbCainLydqOr0ZA5","https://ap.wps.com/l/cbCainLydqOr0ZA5","pdf",1323989,"English","# Abstract\n# Introduction\n# Related work\n## Implicit hate speech taxonomy","[{\"question\":\"What problem does the paper address?\",\"answer\":\"The paper targets implicit hate speech (im-HS) detection, which is difficult for models compared with explicit hate speech due to subtler encoding strategies.\"},{\"question\":\"What are codetypes in this work?\",\"answer\":\"Codetypes are six encoding strategies defined in the paper as rhetorical tactics extracted from im-HS, used to enhance detection.\"},{\"question\":\"How are codetypes integrated into detection systems?\",\"answer\":\"The paper presents two approaches: prompting large language models to classify sentences from generated responses, and using LLMs as encoders with codetypes embedded during encoding.\"}]","Cracking the Code - Enhancing Implicit Hate Speech Detection through Coding Classification | PDF",38]