展示HN:埃普斯坦文件的全文搜索引擎(OCR和OpenSearch)
嗨,HN,
和很多人一样,我对发布的爱泼斯坦/麦克斯韦法庭文件感到沮丧,因为它们大多数是没有文本层的扫描图像(PDF),这使得无法进行Ctrl+F搜索或程序化分析。
我使用Python、Tesseract和OpenSearch构建了一个处理流程来解决这个问题。
网站: [https://epsteinfilez.com](https://epsteinfilez.com)
技术栈:
- 数据处理:使用ocrmypdf(Tesseract)的Python工作进程对原始文件进行并行OCR处理。
- 搜索:使用OpenSearch对提取的文本进行索引。
- 前端:使用Next.js(SSR)构建用户界面。
- 基础设施:自托管的Docker集群。
功能:
- 在大约15,000页上实现亚秒级全文搜索。
- 在PDF页面上直接高亮搜索词。
- 深度链接到特定页面/文档。
这是一款透明度工具,而非政治工具。我希望让原始的主要来源对研究人员和记者可访问。
欢迎对搜索相关性或索引流程提供反馈!
查看原文
Hi HN,<p>Like many people, I was frustrated that the released Epstein/Maxwell court documents were mostly scanned images (PDFs) with no text layer. This made them impossible to Ctrl+F or analyze programmatically.<p>I built a pipeline to fix this using Python, Tesseract, and OpenSearch.<p>The Site: <a href="https://epsteinfilez.com" rel="nofollow">https://epsteinfilez.com</a><p>The Stack:<p>Ingestion: Python workers using ocrmypdf (Tesseract) to perform parallel OCR on raw files.<p>Search: OpenSearch for indexing the extracted text.<p>Frontend: Next.js (SSR) for the UI.<p>Infrastructure: Self-hosted Docker swarm.<p>Features:<p>Sub-second full-text search across ~15,000 pages.<p>Highlights search terms directly on the PDF page.<p>Deep linking to specific pages/documents.<p>This is a transparency tool, not a political one. I wanted to make the raw primary sources accessible to researchers and journalists.<p>Feedback on the search relevance or indexing pipeline is welcome!