Oracle Bone Glyphs: Parametric Vector Dataset (v0.3)

Label and metadata release accompanying the manuscript An Interpretable Parametric Representation and Open Dataset for Oracle Bone Script: Adaptive Arc Segmentation and SVG Resource (npj Heritage Science, 2026; accepted).

The polyline shape data is byte-for-byte unchanged from v0.2 (verified by hash); all changes in this release are to labels and metadata.

What changed in v0.3

• id renamed glyph_id — the sole identifier of a glyph form, corresponding to the source-library entry number (Shuowen Jiezi ordering).
• Pinyin no longer serves as a character identity. In v0.2, 202 entries carried a pinyin string in place of a character; because pinyin is not injective, this had collapsed distinct entries onto one label (six glyph forms were all labelled yu, six xiang, five yi).
• Of those 202 entries, 200 now carry their exact modern character and Unicode code point; 2 are recorded as unidentified rather than given an approximate label.
• Unicode coverage 1,001 → 1,201 of 1,203. The character set now includes CJK Extension A and one supplementary-plane character (𠕋 U+2054B).
• New fields: modern_char, char_form, encoding_status, pinyin.
• New file provenance_manifest.csv — per-entry identity confidence, the basis of each assignment, source composition notes, number of alternative source forms, and the v0.2 label.
• SVG manifests re-issued with the same schema; the SVG files themselves are unchanged.
• v0.2 files retained as *_v0_2.csv / *_v0_2.json.

See CHANGELOG.md for the full history.

Contents

| | | |---|---| | Glyph forms | 1,203 | | Parametric / label data | jgw_1203_labeled.csv, jgw_1203_labeled.json | | Provenance | provenance_manifest.csv (one row per glyph) | | SVG glyph sets | svg_polyline/ (1,203) and svg_parametric/ (1,203), with manifests | | Pipeline | fitting, adaptive arc segmentation, evaluation and analysis code under src/ |

Encoding. All files are UTF-8 and include CJK Extension A and one supplementary-plane character. Read as UTF-8 and index by code point, not by UTF-16 code unit.

Licences

• Data and figures (/data, /figs) — CC BY 4.0
• Code (/src) — MIT

Citation

Qing-sheng Li and Yu-lin Bian. An Interpretable Parametric Representation and Open Dataset for Oracle Bone Script: Adaptive Arc Segmentation and SVG Resource. npj Heritage Science (2026).