RESON v5.9: A Multi-Layer Computational Framework for Structural Characterization of Writing Systems, Applied to the Voynich Manuscript

This deposit contains the complete reproducibility package for RESON (Recursive Extraction of Structural and Organizational Networks), a computational framework comprising nine complementary structural metrics for the characterization of writing systems. Applied to the Voynich manuscript (Beinecke MS 408), the framework identifies a statistically well-defined Intermediate Structural Regime (ISR) that no tested generation mechanism reproduces. Key results (v5.9): A 47-morpheme EBNF grammar achieves 92% token coverage (z = +8.03 vs 18 natural languages, Bonferroni-corrected) 11 generation mechanisms tested; none reproduces the joint 9-metric profile Cross-text validation on 6 external corpora confirms grammar manuscript-specificity (46% self vs 0-2.3% on all others) 3 defense experiments: BPE unsupervised (v54), neural baseline (v55), expanded typology (v56) 57 experiments total, 18 control corpora, 4 rounds of hostile panel validation This deposit supersedes v3.0 (DOI: 10.5281/zenodo.19150369) with 21 additional experiments, cross-text validation, defense experiments, expanded bibliography (56 references), and the RESON CLI universal analysis tool. Contents: Paper v5.1 (submission-ready, .docx), RESON CLI (reson_cli.py), Gold standard corpus (voynich_eva.txt), 50+ control corpora, External validation corpora, 57 experiment scripts, OFFICIAL_NUMBERS v5.9, Strategic roadmap and submission package. Version history: v2 (March 2026, "RESON Framework — Computational-Acoustic Analysis") was an early exploratory draft that included speculative acoustic correlates. From v3.0 onward the framework was redefined as purely structural and all acoustic claims were removed. v5.9 supersedes all previous versions in scope, methodology, and results; earlier versions are retained for archival transparency only.本数据集包含RESON（结构与组织网络递归提取，Recursive Extraction of Structural and Organizational Networks）的完整可复现研究套件。该框架是一套包含9种互补结构度量的计算框架，用于刻画书写系统的特征。将该框架应用于伏尼契手稿（Beinecke MS 408）后，研究识别出一个统计学上定义明确的中间结构状态（Intermediate Structural Regime, ISR），且目前所有测试过的生成机制均无法复现该状态。 关键结果（v5.9版本）： 1. 一套包含47个语素的扩展巴科斯-瑙尔范式（Extended Backus-Naur Form, EBNF）语法实现了92%的Token覆盖率（相较于18种自然语言，z值为+8.03，经邦费罗尼校正） 2. 共测试11种生成机制，均无法复现该9项指标的联合分布特征 3. 对6个外部语料库开展跨文本验证，证实该语法具有手稿特异性（手稿自身匹配度为46%，其余所有语料库的匹配度仅为0%~2.3%） 4. 开展3项防御性实验：无监督字节对编码（Byte Pair Encoding, BPE）实验（版本v54）、神经基线实验（版本v55）、扩展类型学实验（版本v56） 5. 总计开展57项实验，包含18个对照语料库，完成4轮严苛专家评审验证 本数据集取代了v3.0版本（DOI：10.5281/zenodo.19150369），新增了21项实验、跨文本验证模块、防御性实验、扩展参考文献列表（共56篇文献）以及RESON命令行界面（Command Line Interface, CLI）通用分析工具。 数据集包含内容：v5.1版可投稿就绪学术论文（.docx格式）、RESON CLI工具（reson_cli.py）、金标准语料库（voynich_eva.txt）、50余个对照语料库、外部验证语料库、57项实验脚本、v5.9版官方编号文件、研究战略路线图与投稿包。