Papers › Reading Order Matters: Information Extraction from Visually-rich Documents by Token...

Reading Order Matters: Information Extraction from Visually-rich Documents by Token Path Prediction

17 Oct 2023arXiv:2310.11016archive 2025-07-28

Chong Zhang, Ya Guo, Yi Tu, Huan Chen, Jinyang Tang, Huijia Zhu, Qi Zhang, Tao Gui

Recent advances in multimodal pre-trained models have significantly improved information extraction from visually-rich documents (VrDs), in which named entity recognition (NER) is treated as a sequence-labeling task of predicting the BIO entity tags for tokens, following the typical setting of NLP. However, BIO-tagging scheme relies on the correct order of model inputs, which is not guaranteed in real-world NER on scanned VrDs where text are recognized and arranged by OCR systems. Such reading order issue hinders the accurate marking of entities by BIO-tagging scheme, making it impossible for sequence-labeling methods to predict correct named entities. To address the reading order issue, we introduce Token Path Prediction (TPP), a simple prediction head to predict entity mentions as token sequences within documents. Alternative to token classification, TPP models the document layout as a complete directed graph of tokens, and predicts token paths within the graph as entities. For better evaluation of VrD-NER systems, we also propose two revised benchmark datasets of NER on scanned documents which can reflect real-world scenarios. Experiment results demonstrate the effectiveness of our method, and suggest its potential to be a universal solution to various information extraction tasks on documents.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

chongzhangfdu/tpp officialmentioned in paper report
WinterShiver/Token-Path-Prediction mentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Entity LinkingKey Information ExtractionKey-value Pair ExtractionNERNamed Entity RecognitionNamed Entity Recognition (NER)Reading Order DetectionRelation ExtractionSemantic entity labelingSentence OrderingToken Classificationnamed-entity-recognitiontoken-classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Entity Linking FUNSD TPP (LayoutMask) F1 79.20 #3 of 7 Archive leaderboard report
Key Information Extraction CORD TPP (LayoutMask) F1 96.92 #6 of 9 Archive leaderboard report
Key-value Pair Extraction RFUND-EN TPP (LayoutLMv3_base) key-value pair F1 50.27 #11 of 13 Archive leaderboard report
Named Entity Recognition (NER) CORD-r TPP (LayoutLMv3) F1 91.85 #1 of 4 Archive leaderboard report
Named Entity Recognition (NER) CORD-r TPP (LayoutMask) F1 89.34 #2 of 4 Archive leaderboard report
Named Entity Recognition (NER) FUNSD-r TPP (LayoutLMv3) F1 80.40 #1 of 4 Archive leaderboard report
Named Entity Recognition (NER) FUNSD-r TPP (LayoutMask) F1 78.19 #3 of 4 Archive leaderboard report
Reading Order Detection ROOR TPP (LayoutLMv3-base) Segment-level F1 42.96 #3 of 4 Archive leaderboard report
Reading Order Detection ReadingBank TPP (LayoutMask) Average Page-level BLEU 98.16 #1 of 2 Archive leaderboard report
Reading Order Detection ReadingBank TPP (LayoutMask) Average Relative Distance (ARD) 0.37 #1 of 2 Archive leaderboard report
Relation Extraction FUNSD TPP (LayoutMask) F1 79.20 #6 of 9 Archive leaderboard report
Semantic entity labeling FUNSD TPP (LayoutMask) F1 85.16 #11 of 15 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections