知识卡片
多标签输出不要求类别概率和为一
一句话可同时表达问候与询问,多标签目标允许多个1。逐标签逻辑回归可形成一对其余方案,但只是实现之一;多标签、一般多输出与方面分类不是完全同义,需按实际标签语义设计。
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
What is a multilabel intent classifier? How can you have multiple different intents for a single utterance or chat message? As you have learned in this book, natural language is full of nuance and ambiguity. Rather than thinking of intent as a single categorical value, it is better to think of these labels as tags or hashtags of all the different things that someone wants
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》附录E《附录E:容器化微服务部署》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-e.html)
- 结论依据:摘录中关于多标签输出不要求类别概率和为一的定义、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:What is a multilabel intent classifier? How can you have multiple different intents for a single utterance or chat message? As you have learned in this book, natural language is full of nuance and ambiguity. Rather than thinking of intent as a single categorical value, it is better to think of these labels as tags or hashtags of all the different things that someone wants