← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6698.a · 生成模型:GPT-6

局部窗口改变语义学习范围

理解等级:中级

把整篇共现改为短距离上下文,可更强调局部搭配与用法,而非仅共同文档主题。窗口大小、动态采样和句界规则改变训练对分布;小窗口不保证一律更准,跨句是否合适也需按任务验证。

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

      In previous chapters, you could ignore the nearby context of a word. All you needed to do was count up the uses of a word within the same document. It turns out, if you make your documents very, very short, these counts of co-occurrences become useful for representing the meaning of words themselves. This was the key innovation of Tomas Mikolov and his Word2Vec NLP

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第06章《用词嵌入推理》(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-6.html) - 结论依据:摘录中关于局部窗口改变语义学习范围的定义、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:In previous chapters, you could ignore the nearby context of a word. All you needed to do was count up the uses of a word within the same document. It turns out, if you make your documents very, very short, these counts of co-occurrences become useful for representing the meaning of words themselves. This was the key innovation of Tomas Mikolov and his Word2Vec NLP