From PDF to RAG-Ready: Evaluating Document Conversion Frameworks for Domain-Specific Question Answering
从PDF到RAG就绪:评估面向特定领域问答的文档转换框架
机构 * Faculty of Engineering, University of Porto(葡萄牙波尔图大学工程学院) ; Department of Business Administration, University of Maia(马亚大学商业管理系) ; LIACC—Artificial Intelligence and Computer Science Laboratory, University of Porto(葡萄牙波尔图大学人工智能与计算机科学实验室) ; Department of Communication Sciences and Information Technologies, University of Maia(马亚大学通讯科学与信息科技系) ; School of Health, Polytechnic of Porto(波尔图理工学院健康学院) ; School of Technology and Management, Polytechnic Institute of Maia(马亚理工学院技术与管理学院)
专题命中 知识库问答 :RAG(title,title_cn);retrieval-augmented generation(abstract);分类 cs.IR、cs.AI
AI总结 通过系统比较四种开源PDF转Markdown框架的21种流水线配置,发现文档预处理质量(尤其是层次化分块和元数据增强)对RAG系统问答准确率的影响远超转换工具本身,最佳配置(Docling+层次化分块+图像描述)达到94.1%准确率,超越人工整理。
Comments 27 pages, 3 figures, 7 tables
Journal ref Applied Sciences 16 (2026) 5069