An interpretable parametric representation and open dataset for oracle bone script: adaptive Arc segmentation and SVG resource

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Computational oracle bone script (OBS) research is dominated by opaque, data-hungry bitmap-domain deep models. We present an interpretable parametric representation and an open dataset of 1203 oracle-bone glyph forms in which each stroke sub-segment is encoded by three closed-form parameters—direction α, magnitude η, signed curvature κ— each entry carrying a modern-character assignment and, where that modern character is encoded, a Unicode code point, an error-driven adaptive arc-segmentation method via residual-guided splitting, corpus-level structural characterisation, a controllability demonstration, and a resolution-independent SVG glyph set. The representation achieves a median normalised reconstruction error of 0.0021 at ~10 sub-segments per glyph (4× lower than a single-arc baseline); 72.0% of sub-segments turn through less than 0.16 rad, quantifying, for this digitised corpus, the palaeographic view that OBS strokes are simple low-curvature arcs. The release is a single-representative-glyph-per-character vector resource—a complement to primary palaeographic sources, not a substitute.