知识卡片
对数压缩频率比不使分布均匀
取对数可压缩数量级,例如以10为底时一百万和十万分别变为6和5。这不把任意频率分布变成均匀分布,齐普夫幂律也不是指数律;在没有外加常数的整体尺度变换中,对数底改变只带来统一倍数。
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
3.6.1 Return of Zipf You’re almost there. Let’s say, though, you have a corpus of one million documents (maybe you’re baby Google) and someone searches for the word cat. In your one million documents, suppose you have exactly one document that contains the word cat. The raw IDF of this is 1,000,000 / 1 = 1,000,000 Let’s imagine you have 10 documents with the word
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第03章《词频与逆文档频率向量》(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-3.html)
- 结论依据:摘录中关于对数压缩频率比不使分布均匀的定义、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:3.6.1 Return of Zipf You’re almost there. Let’s say, though, you have a corpus of one million documents (maybe you’re baby Google) and someone searches for the word cat. In your one million documents, suppose you have exactly one document that contains the word cat. The raw IDF of this is 1,000,000 / 1 = 1,000,000 Let’s imagine you have 10 documents with the word