<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[PDF 表格该直接转 Markdown，还是单独建表？]]></title><description><![CDATA[<p dir="auto">供应商报价 PDF 里很多表格。直接转 Markdown 喂给 RAG，回答数字经常错位。</p>
]]></description><link>https://localaihub.com/topic/62/pdf-表格该直接转-markdown-还是单独建表</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 19:23:42 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/62.rss" rel="self" type="application/rss+xml"/><pubDate>Sun, 03 May 2026 06:53:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Mon, 04 May 2026 05:47:00 GMT]]></title><description><![CDATA[<p dir="auto">这句话可以贴墙上。拿 PDF 当唯一事实源，后面会一直补洞。</p>
]]></description><link>https://localaihub.com/post/213</link><guid isPermaLink="true">https://localaihub.com/post/213</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Mon, 04 May 2026 05:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Mon, 04 May 2026 03:47:00 GMT]]></title><description><![CDATA[<p dir="auto">对。PDF 是交付格式，不一定是知识库最佳原始格式。</p>
]]></description><link>https://localaihub.com/post/212</link><guid isPermaLink="true">https://localaihub.com/post/212</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Mon, 04 May 2026 03:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Mon, 04 May 2026 01:51:00 GMT]]></title><description><![CDATA[<p dir="auto">我先把报价表拆成结构化 CSV，再让 RAG 引用 CSV 行，不直接让模型读 PDF 表。</p>
]]></description><link>https://localaihub.com/post/211</link><guid isPermaLink="true">https://localaihub.com/post/211</guid><dc:creator><![CDATA[小蓝]]></dc:creator><pubDate>Mon, 04 May 2026 01:51:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Mon, 04 May 2026 01:32:00 GMT]]></title><description><![CDATA[<p dir="auto">这个折中不错，但要防摘要漏列。价格、阈值、日期别只在摘要里。</p>
]]></description><link>https://localaihub.com/post/210</link><guid isPermaLink="true">https://localaihub.com/post/210</guid><dc:creator><![CDATA[小路灯]]></dc:creator><pubDate>Mon, 04 May 2026 01:32:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 22:54:00 GMT]]></title><description><![CDATA[<p dir="auto">我们做过一个折中：表格摘要进向量，原始 CSV 进对象存储，回答时附表格片段。</p>
]]></description><link>https://localaihub.com/post/209</link><guid isPermaLink="true">https://localaihub.com/post/209</guid><dc:creator><![CDATA[米饭]]></dc:creator><pubDate>Sun, 03 May 2026 22:54:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 21:39:00 GMT]]></title><description><![CDATA[<p dir="auto">是的，RAG 不等于所有东西都塞向量库。数字类事实经常更适合 SQL。</p>
]]></description><link>https://localaihub.com/post/208</link><guid isPermaLink="true">https://localaihub.com/post/208</guid><dc:creator><![CDATA[MingK]]></dc:creator><pubDate>Sun, 03 May 2026 21:39:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 21:17:00 GMT]]></title><description><![CDATA[<p dir="auto">那是不是知识库要连数据库？</p>
]]></description><link>https://localaihub.com/post/207</link><guid isPermaLink="true">https://localaihub.com/post/207</guid><dc:creator><![CDATA[小树]]></dc:creator><pubDate>Sun, 03 May 2026 21:17:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 19:21:00 GMT]]></title><description><![CDATA[<p dir="auto">如果表格是核心事实，建议另建结构化表，RAG 只负责找相关表和解释口径。</p>
]]></description><link>https://localaihub.com/post/206</link><guid isPermaLink="true">https://localaihub.com/post/206</guid><dc:creator><![CDATA[阿航]]></dc:creator><pubDate>Sun, 03 May 2026 19:21:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 16:56:00 GMT]]></title><description><![CDATA[<p dir="auto">至少给每行生成 row_id，引用时能回到“第几页第几行”，不然审计没法做。</p>
]]></description><link>https://localaihub.com/post/205</link><guid isPermaLink="true">https://localaihub.com/post/205</guid><dc:creator><![CDATA[rootless]]></dc:creator><pubDate>Sun, 03 May 2026 16:56:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 16:02:00 GMT]]></title><description><![CDATA[<p dir="auto">这类应该检索到表格后走计算或表格问答，不要让模型凭文本猜坐标。</p>
]]></description><link>https://localaihub.com/post/204</link><guid isPermaLink="true">https://localaihub.com/post/204</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Sun, 03 May 2026 16:02:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 14:25:00 GMT]]></title><description><![CDATA[<p dir="auto">我们把表格转 Markdown 后，模型会把上一行的价格拿给下一行产品。</p>
]]></description><link>https://localaihub.com/post/203</link><guid isPermaLink="true">https://localaihub.com/post/203</guid><dc:creator><![CDATA[青菜]]></dc:creator><pubDate>Sun, 03 May 2026 14:25:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 12:49:00 GMT]]></title><description><![CDATA[<p dir="auto">PyMuPDF 也有表格提取能力，可以作为候选。关键是要人工验几类 PDF。</p>
]]></description><link>https://localaihub.com/post/202</link><guid isPermaLink="true">https://localaihub.com/post/202</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Sun, 03 May 2026 12:49:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 11:31:00 GMT]]></title><description><![CDATA[<p dir="auto">Camelot 对规整表格不错，但扫描件和复杂合并单元格会翻车。</p>
]]></description><link>https://localaihub.com/post/201</link><guid isPermaLink="true">https://localaihub.com/post/201</guid><dc:creator><![CDATA[小高]]></dc:creator><pubDate>Sun, 03 May 2026 11:31:00 GMT</pubDate></item><item><title><![CDATA[Reply to PDF 表格该直接转 Markdown，还是单独建表？ on Sun, 03 May 2026 09:04:00 GMT]]></title><description><![CDATA[<p dir="auto">报价表不要只当文本。列名、单位、币种、有效期都要结构化。</p>
]]></description><link>https://localaihub.com/post/200</link><guid isPermaLink="true">https://localaihub.com/post/200</guid><dc:creator><![CDATA[MingK]]></dc:creator><pubDate>Sun, 03 May 2026 09:04:00 GMT</pubDate></item></channel></rss>