Toward a Phonetic Decryption of the Voynich Manuscript: A Human-AI Collaborative Pipeline Using King-Andrisani Transliteration and Agglutinative Segmentation

Toward a Phonetic Decryption of the Voynich Manuscript (Beinecke MS 408) A computational pipeline (K&A v12) for Latin phonetic recovery from the Voynich Manuscript, achieving 89.3% dictionary validation across 38,442 decoded words on 226 folio sides. Key results: 90.6% high-confidence Latin recovery using monolithic-first decoding + agglutinative prefix detection (13 rules) 33 pharmaceutical terms identified, including ingredients on astronomical pages (nardi, cassiae, apii, vini) Folio f57v identified as a medico-astrological volvelle, structurally isomorphic to Ashmole 370 (~1424) 7/12 Aurea Alexandrina ingredients matched on folio f103r (Antidotarium Nicolai cross-validation) Dual cipher system demonstrated: phonetic substitution + nomenclator codebook This deposit contains: The full paper (52 pages, including a 226-folio catalogue with page thumbnails) The paper is accompanied by a public GitHub repository (github.com/CorwinFr/Voynich) containing the complete pipeline source code (Python), decoded Latin text for all folios, and an enriched folio-by-folio JSON catalogue. Authors: Guillaume Clement (Flow Line Integration) assisted by Claude (Anthropic, Opus 4.6). Registered via Enveloppe Soleau, INPI (France), April 2026.