← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6674 · 生成模型:GPT-6

主题坐标是词项的加权组合

理解等级:中级

将词频或TF–IDF向量乘以一个词项到主题的权重矩阵,可在较少坐标中表示文档。某一主题可同时接受多个词的贡献,正负权重也可相互抵消;这些坐标反映拟合结构,不自动是命名清楚、可判真假的概念。

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

      You’d need to find those word dimensions that “belong” together in a topic and add their TF–IDF values together to create a new number to represent the amount of that topic in a document. You might even weight them for how important they are to the topic, how much you’d like each word to contribute to the “mix.” And you could have negative weights for

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第04章《从词数到语义分析》(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-4.html) - 结论依据:摘录中关于主题坐标是词项的加权组合的定义、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:You’d need to find those word dimensions that “belong” together in a topic and add their TF–IDF values together to create a new number to represent the amount of that topic in a document. You might even weight them for how important they are to the topic, how much you’d like each word to contribute to the “mix.” And you could have negative weights for