将文档中的图文等元素整合为语义完整单元，提升检索准确率。
Evidence Units: Ontology-Grounded Document Organization for Parser-Independent Retrieval
- 基于本体构建统一语义框架，融合不同解析器输出。
- 通过全局相似度匹配实现段落与图文的最优归组。
- 在跨解析器场景下保持效果稳定，适合学术文档检索。
结构化文档（如表格与标题、图表与说明、公式与解释段落）在检索索引时常被碎片化。现有元素级索引将每个解析出的元素视为独立片段，导致语义完整的单元分散在多个候选结果中。本文提出一种无需依赖特定解析器的流程，构建证据单元（Evidence Units, EUs）：将视觉元素与其上下文文本组合成语义完整的文档块。主要贡献包括：(1) 基于本体的角色标准化扩展DoCO，将异构解析器输出映射至统一语义模式；(2) 基于全相似度矩阵的全局段落分配算法，实现段落与EUs的最优匹配；(3) 在Neo4j中采用图结构决策层，形式化构建规则并用两个不变量验证完整性；(4) 跨解析器验证显示，即使在MinerU与Docling之间边界框差异下，证据单元的空间分布仍趋同，性能增益保持稳定。在OmniDocBench v1.0（1,340页，1,551个问答对）上的实验表明，基于EUs的分块使检索长序列一致性（LCS）提升+0.31（从0.50到0.81），召回率@1从0.15升至0.51（提升3.4倍），最小排名（MinK）从2.58降至1.72。跨解析器结果确认，性能增益在不同解析器间稳定保持（LCS提升+0.23至+0.31）。文本查询表现最显著：召回率@1从0.08跃升至0.47。
原文摘要 · Abstract (English)
Structured documents--tables paired with captions, figures with explanations, equations with the paragraphs that interpret them--are routinely fragmented when indexed for retrieval. Element-level indexing treats every parsed element as an independent chunk, scattering semantically cohesive units across separate retrieval candidates. This paper presents a parser-independent pipeline that constructs Evidence Units (EUs): semantically complete document chunks that group visual assets with their contextual text. We introduce four contributions: (1) ontology-grounded role normalization extending DoCO that maps heterogeneous parser outputs to a unified semantic schema; (2) a semantic global assignment algorithm that optimally assigns paragraphs to EUs via a full similarity matrix; (3) a graph-based decision layer in Neo4j that formalizes EU construction rules and validates completeness through two invariants; and (4) cross-parser validation showing EU spatial footprints converge across MinerU and Docling, with gains preserved under parser-induced bbox variance. Experiments on OmniDocBench v1.0 (1,340 pages; 1,551 QA pairs) show EU-based chunking improves retrieval LCS by +0.31 (0.50 to 0.81). Recall@1 increases from 0.15 to 0.51 (3.4x) and MinK decreases from 2.58 to 1.72. Cross-parser results confirm the gain (LCS +0.23 to +0.31) is preserved across parsers. Text queries show the most dramatic gain: Recall@1 rises from 0.08 to 0.47.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。