Structural Vocabulary Analysis of the Voynich Manuscript: Jensen-Shannon Divergence Matrix, Spectral Clustering, and Dimensional Entropy Signatures

Structural analysis of the Voynich manuscript's five-section vocabulary distribution using Jensen-Shannon divergence, spectral clustering, and per-section conditional entropy. The JSD matrix reveals a significantly stretched distribution (ratio=1.373, p<0.0001 vs 5,000 permutations). Spectral clustering with k=3 recovers exactly the three predicted dimensional clusters (O/R/α), stable across eight random seeds (ARI=1.000). Conditional entropy analysis reveals a dimensional split: Biological section has the lowest character-level entropy (h₂=1.990) and highest word-level entropy, while Astronomical has the highest character-level entropy (h₂=2.440) and lowest word-level entropy—consistent with void framework predictions about dimensional encoding structure.