This repository contains the paper, analysis scripts, and data demonstrating that Boltzmann's formula S = k ln(W) applies to natural language at four orthogonal levels. Using WordNet on 100 concepts, four types of W (expressive, semantic, ambiguity, synonymy) are defined and shown to be statistically independent (all pairwise ρ < 0.06). An initial correlation with taxonomic relations was identified as structurally inflated; an independent Wikipedia proxy test is reported as an honest negative result.
Massimo Lacchè (Fri,) studied this question.