知识卡片
相似度数值不是语义百分比
余弦相似度0.117只是给定特征空间和权重下的方向接近程度,不能解释为共享11.7%的意义。分词、词表、归一化和领域都会改变数值;阈值应在具体相关性标注上校准,而非把数字自带当概率。
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
This count vector for the first sentence from the “Algorithmic Bias” article is only 11.7% similar (cosine similarity of 0.117) to the second sentence of the article. It seems that the second sentence shares very few words with the first sentence.
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:摘录中关于相似度数值不是语义百分比的定义、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:This count vector for the first sentence from the “Algorithmic Bias” article is only 11.7% similar (cosine similarity of 0.117) to the second sentence of the article. It seems that the second sentence shares very few words with the first sentence.