Version 3 of the candidate decipherment of the Voynich Manuscript, identifying it as a 15th-century Sri Lankan Elu-Sinhala pharmaceutical text. This version supersedes V2 (Zenodo DOI 10.5281/zenodo.20023733, 2026-05-04). Major new findings in V3: V20 corpus expanded to 36,633 tokens (NL-merger); coverage raised to 92.9% fully resolved (was 52–65% in V2 under old compound-splitter methodology); VPNS expanded to 21 base preparation states (was 8/18); 25 BM score-3+ formula clusters across 47 lines (was 4); nura=nirūha (Panchakarma herbal enema) confirmed by two BM attestations; BM Chapter 57 piśāca+nāga textual parallel for f113r Bhūta-cikitsā chapter; 23 RECIPE therapeutic categories mapped with three-tier visual margin system confirmed by Beinecke IIIF inspection; 18/23 opener words identified as Elu compressions of Sanskrit chapter-title vocabulary; recipe chapter sequence mapped to AH+BM hybrid Sri Lankan ordering; COSMO section quantified as temporal prescription index (10–15% cross-reference rate to all sections; 90–100% preparation-vocabulary coherence; 3 nakshatras confirmed: acena=Aśvinī, alula=Anurādhā, makhea=Maghā); ASTRO section reidentified as timed pharmaceutical processing schedule (not astronomical); Rosette page identified as master preparation map (teda 3.9× peak); herbal visual matching: 92.2% of 110 folios show visual features matching decoded vocabulary; 5 VPNS state-markers now strongly externally grounded (was 2). P(Sinhala identification wrong) ≈ 7–10%. Licensing: ORIGINAL CONTENT (paper, code, analysis writeups) is CC-BY-4.0. Bundled third-party reference corpora retain their licenses (GRETIL CC-BY-NC-SA; Wellcome and thesis materials under research fair-use). See LICENSE file in deposit.
Kameldip Singh Basra (Thu,) studied this question.