Ernie.SG Semantic Responsive Typesetting
The page is a rendition, not the document.
Abstract. We investigate whether one semantic scholarly document can produce coherent renditions for print, e-ink, and continuous reading while preserving identity, relationships, reading position, and annotations.
1. Proposition
Documents are usually flattened into pages too early. Once meaning becomes coordinates, every new screen or sheet becomes a repair job. Semantic Responsive Typesetting reverses that relationship: the document stays persistent; pages, columns, and line breaks are computed for a target.
One semantic document. Multiple computed renditions. Persistent identity, relationships, reading position, and annotations.
2. Composition model
A canonical graph is combined with a target profile, a deterministic composition policy, and optional target-specific overrides. The renderer selects component variants, measures flow, records violations, and emits both the rendition and an inspectable manifest.
3. Apparent research gap
Adaptive layout, PDF reflow, scholarly reconstruction, and robust annotation anchoring each have substantial prior art. The credible contribution is their integration into a fidelity-constrained, finite-height system and its evaluation across materially different reading surfaces.
4. First experiment
The first experiment holds content constant and varies representation geometry: A4 print, reMarkable Paper Pro, reMarkable Paper Pro Move, and continuous mobile. We test structural preservation, layout quality, anchor continuity, and composition time.