Reproduce “From Historical Tabular Image to Knowledge Graphs: A Provenance-Aware Modular Pipeline” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sarah Binta Alam Shoilee, Victor de Boer, Jacco van Ossenbruggen, Susan Legêne
A modular, provenance-aware pipeline is introduced for constructing knowledge graphs from historical handwritten tabular document images, enabling explicit traceability of extracted assertions to their visual and textual origins. The approach facilitates systematic error analysis across stages and underscores the importance of human intervention for reliable information extraction from complex, degraded archival materials.
Key insights
- Modularity enables precise diagnosis and attribution of errors (e.g., HTR vs. TSR issues), revealing hidden flaws in complex models that would be obscured in end-to-end systems.
- While provenance ensures traceability, it does not guarantee semantic correctness; the reliability of upstream components (Handwritten Text Recognition, Table Structure Recognition) is critical for accurate downstream results.
- For low-quality, complex historical documents, human intervention for structural correction and semantic refinement is essential, demonstrating the necessity of human-AI collaboration for reliable KG construction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.