知识卡片
人工监督的成本与覆盖取舍
作者的招聘辅助案例同时保留抽样复查和低置信项人工处理,结果并未节省时间或金钱,却扩大了候选覆盖。这说明自动化收益不只有替代成本,也可能改变可处理范围;它只是作者案例,不是以预测分数决定招聘的通用方案。
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
原书出处 · 1 处
- · 查看原书依据
NLP pipelines, like all algorithms, make mistakes and are almost always biased in many ways, so if you use NLP to automate communication with humans, be careful. At Tangible AI, where I am the CTO, we used NLP for the critical business process of finding developers to join our team, so we supervised our NLP pipeline carefully. NLP was allowed to filter out job applications
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第01章《会读写的机器与自然语言处理》(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-1.html)
- 结论依据:摘录中关于人工监督的成本与覆盖取舍的定义、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:NLP pipelines, like all algorithms, make mistakes and are almost always biased in many ways, so if you use NLP to automate communication with humans, be careful. At Tangible AI, where I am the CTO, we used NLP for the critical business process of finding developers to join our team, so we supervised our NLP pipeline carefully. NLP was allowed to filter out job applications