A Categorical Naive Bayes Model of Morphological Slot Dynamics in the Voynich Manuscript: From Intuition to a Low-Entropy Attribute-Tagging Architecture

Abstract

The Voynich Manuscript (MS 408) has long resisted decipherment under standard natural language translation paradigms and classical substitution ciphers. In this paper, we document the complete trajectory—from foundational intuition to a mathematically formalized computational model—of reframing MS 408 not as an unmapped spoken language, but as a synthetic, low-entropy attribute-tagging system operating under strict phonotactic constraints. We propose a 16-slot morphological decomposition framework (S0 . . . S15) coupled with a Laplace-smoothed (α = 1.0) Categorical Naive Bayes classifier. Evaluated across canonical manuscript folios—including an in-depth empirical transition test on f84r—our model demonstrates robust domain discrimination, effectively isolating active functional domains (yielding a 50.64% Pharmaceutical / 49.36% Balneological split with H = 0.4999) while completely rejecting non-relevant domains (0.00% for Botanical and Astronomical). These quantitative findings mathematically refute the medieval hoax hypothesis, proving the existence of a coherent, domain-adaptive generative grammar. All mathematical formalizations, log-likelihood transformations, and code generation routines were executed via artificial intelligence collaboration.

Keywords