针对游戏聊天中的毒性内容，提出一种兼顾语言与符号的检测新方法。
thaulab@EEUCA 2026: Who Said What to Whom? A Targeting-Aware Neural-Symbolic Pipeline for Gaming Toxicity Detection
- 用双模型融合加语言中介模块处理分歧，提升判断准确性。
- 在极端不平衡数据下实现0.6441的宏平均F1和0.9062准确率。
- 特别关注仇恨、威胁等高危类别，适合游戏安全监控场景。
本文介绍我们在EEUCA 2026共享任务中针对游戏聊天毒性分类的系统。采用三阶段流程：集成两种小型Transformer（DeBERTa-v3-base，184M；XLM-RoBERTa-base，278M），结合语言学引导的中介模块（LIM），通过语料支持的词汇归一化、条件性词频评分、多语言粗俗语检测及基于言语行为理论的主动目标分析，解决模型间分歧。该模块重点优化少数类（仇恨与骚扰、威胁、极端主义），这些类别在实际游戏监管中最为关键。面对1,450:1的极端类别不平衡问题，仅使用原始训练数据设计了两阶段数据增强策略。系统在官方测试集上取得0.6441的宏平均F1和0.9062的准确率，排名宏平均F1第三、准确率第一。该管道具备领域可迁移性：只需替换特定游戏的实体词典即可适配其他平台。代码已公开于https://github.com/Anmol2059/thaulab_EEUCA。
原文摘要 · Abstract (English)
This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat. We implement a three-stage pipeline combining an ensemble of two compact transformers (DeBERTa-v3-base, 184M; XLM-RoBERTa-base, 278M) with a Linguistically-Informed Mediator (LIM) that resolves inter-model disagreements through corpus-backed lexical normalization, class-conditional unigram scoring, multilingual profanity detection, and agentive targeting analysis grounded in speech act theory. The LIM specifically targets the minority classes (Hate \& Harassment, Threats, and Extremism), which are the most safety-critical categories in real-world gaming moderation. To address the extreme class imbalance (1{,}450:1 Non-toxic to Extremism ratio), we introduce a two-stage data augmentation strategy using only the provided training data. Our system achieves a Macro F1 of 0.6441 and accuracy of 0.9062 on the official test set, ranking 3rd in Macro F1 and 1st in accuracy among all teams. The proposed pipeline is domain-portable: adapting to other gaming platforms requires substituting only the game-specific entity lexicon. Code is publicly available at https://github.com/Anmol2059/thaulab\_EEUCA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。