知识卡片

普通读书笔记卡 · 6615 · 生成模型:GPT-6

去重需明确相似类型与粒度

理解等级:中级

完全相同、近似重复和语义重复是不同任务;整篇重复与跨文档片段重复也需不同处理。应先指定相似表示、阈值和删除单位,再解释去重比例,不能用“已去重”概括全部重合风险。

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

      How do we define a duplicate? We will make a distinction between three kinds: Exact matches Two sequences with the same text are exact-match duplicates. They are the easiest to handle. Approximate matches In many cases, there are near-duplicates, where sequences of text are identical except for a few characters. Sometimes these sequences are slightly different only due to HTML text extraction artifacts and other

    参考来源

    - 位置:《Designing Large Language Model Applications》第02章《预训练数据》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch02.html) - 结论依据:摘录中关于去重需明确相似类型与粒度的定义、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:How do we define a duplicate? We will make a distinction between three kinds: Exact matches Two sequences with the same text are exact-match duplicates. They are the easiest to handle. Approximate matches In many cases, there are near-duplicates, where sequences of text are identical except for a few characters. Sometimes these sequences are slightly different only due to HTML text extraction artifacts and other