Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces Evidence Units (EUs), a parser-independent pipeline for constructing semantically complete document chunks by grouping visual assets with their contextual text. They use ontology-grounded role normalization and a semantic global assignment algorithm to map heterogeneous parser outputs to a unified semantic schema, followed by a graph-based decision layer to formalize EU construction rules. Experiments on OmniDocBench demonstrate that EU-based chunking significantly improves retrieval performance, particularly for text queries, with a 3.4x increase in Recall@1.
Stop scattering your structured documents: Evidence Units boost retrieval Recall@1 by 3.4x by intelligently chunking visual assets with their contextual text, regardless of the parser used.
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.