知识卡片
毒性标签是任务约定而非意图证明
评论数据把威胁、侮辱等内容归为毒性,训练系统预测的是这些标注。单条文本不必足以证明作者真实意图,引用或讨论相关词也可能并非攻击;定义与标注边界需要在错例和实际使用场景中持续检查。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 从词数到语义分析
原书出处 · 1 处
- · 查看原书依据
To see the power of topic modeling, you will solve a real problem: recognizing toxicity in Wikipedia comments. A toxic comment is a message that intends to upset, humiliate, or scare whoever it’s addressing. Threats, insults, obscenities, and hate speech are all forms of toxicity. This kind of toxicity can be especially dangerous on reference websites, such as Wikipedia, where toxicity in the dialog among
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第04章《从词数到语义分析》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-4.html)
- 结论依据:摘录中关于毒性标签是任务约定而非意图证明的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:To see the power of topic modeling, you will solve a real problem: recognizing toxicity in Wikipedia comments. A toxic comment is a message that intends to upset, humiliate, or scare whoever it’s addressing. Threats, insults, obscenities, and hate speech are all forms of toxicity. This kind of toxicity can be especially dangerous on reference websites, such as Wikipedia, where toxicity in the dialog among