知识卡片
稠密检索仍可精确但成本需评估
稠密向量多数维非零,直接按非零项建倒排表难以有效缩小候选。精确检索仍可实现,只是高维下常有成本挑战;倒排索引也不要求权重必须整数,关键在稀疏性及可用剪枝条件。
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
- · 从词数到语义分析
原书出处 · 1 处
- · 查看原书依据
But unlike BOW and TF–IDF tables, tables of semantic vectors can’t be easily indexed using traditional inverted index techniques.32 Inverted indexes work for discrete vectors or binary vectors, like tables of binary or integer word document vectors because the index only needs to maintain an entry for each nonzero discrete dimension. Either the value of that dimension is present or it isn’t in the referenced
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:摘录中关于稠密检索仍可精确但成本需评估的定义、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:But unlike BOW and TF–IDF tables, tables of semantic vectors can’t be easily indexed using traditional inverted index techniques.32 Inverted indexes work for discrete vectors or binary vectors, like tables of binary or integer word document vectors because the index only needs to maintain an entry for each nonzero discrete dimension. Either the value of that dimension is present or it isn’t in the referenced