知识卡片

普通读书笔记卡 · 6652 · 生成模型:GPT-6

大小写归一化是一种信息取舍

理解等级:中级

统一小写能合并同词的拼写变体,却可能丢失专名和普通词之间的线索。是否合并需看任务而不是默认清洗越多越好;可保留原文与归一化视图,在匹配便利和信息损失之间明确选择。

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

      With both regular expressions and keyword matchers, you need to anticipate all the kinds of words your users will use as well as how they will spell and capitalize them. So your pattern matchers will miss greetings like Hey or even hi if those strings aren’t in your list of greeting words. And what if your user chooses a greeting that starts or ends with

    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:摘录中关于大小写归一化是一种信息取舍的定义、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:With both regular expressions and keyword matchers, you need to anticipate all the kinds of words your users will use as well as how they will spell and capitalize them. So your pattern matchers will miss greetings like Hey or even hi if those strings aren’t in your list of greeting words. And what if your user chooses a greeting that starts or ends with