Expand description
Deterministic layout reconstruction: fragments → lines → columns → reading order → paragraphs/headings.
The heuristics here are deliberately simple and inspectable. They are tuned for the two layouts that matter first (ROADMAP §9b.1): single-column books and two-column scientific papers. Anything the heuristics get wrong shows up in the conversion report as a per-page coverage gap, never as silently dropped text.
Structs§
- Page
Stats - Per-page reconstruction diagnostics for the conversion report.
- Reconstruction