https://www.hansking.cn/blog A posts created with Next.js and Tailwind.css en-us admin@hansking.cn (HANS) admin@hansking.cn (HANS) Sun, 29 Sep 2024 10:00:00 GMT https://www.hansking.cn/blog/2026/RAG-shu-ju-guan-dao-jie-xi-PDF-wang-ye-yu-biao-ge https://www.hansking.cn/blog/2026/RAG-shu-ju-guan-dao-jie-xi-PDF-wang-ye-yu-biao-ge [{"value":"RAG 数据管道：解析 PDF、网页与表格","url":"#rag-数据管道解析-pdf网页与表格","depth":1},{"value":"一、先承认：不同格式不是同一种文档","url":"#一先承认不同格式不是同一种文档","depth":2},{"value":"二、PDF：先处理布局，再处理文本","url":"#二pdf先处理布局再处理文本","depth":2},{"value":"三、网页：正文只是页面的一小部分","url":"#三网页正文只是页面的一小部分","depth":2},{"value":"四、表格：关系比自然语言更重要","url":"#四表格关系比自然语言更重要","depth":2},{"value":"五、切分：不是越小越好，也不是越大越完整","url":"#五切分不是越小越好也不是越大越完整","depth":2},{"value":"六、清洗要克制：噪声和信息不能一锅端","url":"#六清洗要克制噪声和信息不能一锅端","depth":2},{"value":"七、去重和版本：避免知识库里有十个“同一份答案”","url":"#七去重和版本避免知识库里有十个同一份答案","depth":2},{"value":"八、权限元数据必须跟着每个 chunk 走","url":"#八权限元数据必须跟着每个-chunk-走","depth":2},{"value":"九、质量检查要在入库时就发生","url":"#九质量检查要在入库时就发生","depth":2},{"value":"十、增量处理：文档变化时只做必要工作","url":"#十增量处理文档变化时只做必要工作","depth":2},{"value":"十一、如何评测管道是否真的变好了","url":"#十一如何评测管道是否真的变好了","depth":2},{"value":"十二、我的总结：RAG 的第一模型是数据管道","url":"#十二我的总结rag-的第一模型是数据管道","depth":2}] Sun, 29 Sep 2024 10:00:00 GMT admin@hansking.cn (HANS) AILLMRAG数据工程文档解析