← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8703.e · 生成模型:GPT-6

版面理解补充文本结构线索

理解等级:中级

字体、位置、表格和图文关系可帮助识别标题与作用域;视觉嵌入还可直接检索页面。方法选择应检查信息保留和成本,按页编码也可能漏掉跨页关系,不能宣称完全不再需要任何分块设计。

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

      Layout-Aware Chunking A more involved form of metadata-aware chunking is layout-aware chunking. In this approach we use computer vision techniques to extract layout information about the document, including the placement and scope of textual elements, the titles, subtitles, font size of text, etc.; use this metadata to inform the chunking process. Both open source and proprietary tools can facilitate layout extraction. They include tools like

    参考来源

    - 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html) - 结论依据:摘录中关于版面理解补充文本结构线索的定义、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:Layout-Aware Chunking A more involved form of metadata-aware chunking is layout-aware chunking. In this approach we use computer vision techniques to extract layout information about the document, including the placement and scope of textual elements, the titles, subtitles, font size of text, etc.; use this metadata to inform the chunking process. Both open source and proprietary tools can facilitate layout extraction. They include tools like