Global Term Key: Constitutional Lexical Identity Through Disambiguated Semantic Binding

Language systems routinely perform lexical disambiguation. Computational linguistics performs Word Sense Disambiguation (WSD) to assign meanings to tokens in context. Lexicography groups synonymous meanings into synsets. Knowledge graphs map names to canonical entities. Semantic web systems assign URIs to concepts. These systems answer a practical question: Which meaning is intended here? This paper argues that the practical question, while useful, is incomplete. Disambiguation alone classifies. It does not constitutionalize. This paper introduces Global Term Key (GTK): a constitutional lexical primitive that treats the disambiguated semantic binding as the unit of invariant analysis rather than the undifferentiated word-form. Under the current compact definition, GTK = RPZ1+D: a disambiguated named view that makes a semantic identity reachable from its machine-oriented CSIL locator. Its RPZ1 structure includes a binding-identity reference, a governed alias set, and a competing-binding set for overloaded surface forms. The binding's bounded intension or invariant-floor candidates belong primarily to RPZ2; its wider typed adjacency belongs primarily to RPZ3. GTK links those layers without collapsing locating, naming, defining, and relating meaning into one object. Existing lexical and knowledge technologies provide comparison surfaces for sense selection, synonymy, entity reference, and competing meanings. The paper's proposed constitutional addition is governed resolution, membership, provenance, standing, refusal, and revision across the layered identity-and-naming structure; novelty remains to be externally assessed. The central claim is: Invariant lexical structure attaches not to ambiguous surface forms, but to lexically disambiguated concept-bindings. Polysemy is therefore not evidence against invariant lexical structure. It is evidence that one token indexes multiple GTKs.