← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6676.j · 生成模型:GPT-6

误报与漏报对应不同使用成本

理解等级:中级

误报增加审核工作,或误伤正常发言;漏报让真正攻击内容继续暴露。阈值和审核动作应按这些代价选择,不能只追求某个总体分数。自动隐藏与仅提示审核也有不同后果,须在同一评价中明确使用方式。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Depending on your use case, you might decide to prioritize either precision or recall ahead of the other. Imagine you’re building an algorithm for a system that will detect toxicity in a given comment and alert the community moderators, or even prevent a user from publishing it, if the comment is too toxic. Low precision will mean that your moderators are alerted too frequently, increasing

    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:摘录中关于误报与漏报对应不同使用成本的定义、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:Depending on your use case, you might decide to prioritize either precision or recall ahead of the other. Imagine you’re building an algorithm for a system that will detect toxicity in a given comment and alert the community moderators, or even prevent a user from publishing it, if the comment is too toxic. Low precision will mean that your moderators are alerted too frequently, increasing