Self-Referential Natural-Language Formalization: A Minimal Mathematical Model and MVP Pipeline for Meaning-Generating Dynamics

This paper proposes a computational framework for formalizing natural-language concepts not as static definitions but as executable dynamical objects. A concept is represented as a pair of operators: a lower-layer meaning-generation rule ( f ) and an upper-layer relational update operator ( \Phi ). The lower layer models semantic state transitions as a contractive linear dynamical system, while the upper layer updates a weighted concept-relation matrix according to the current semantic activation pattern. Under iterative application, the system converges to a semantic condensation point, which is interpreted as the essence of the concept. We present the minimal mathematical formulation, an algorithmic pipeline, and a runnable MVP implementation that demonstrates end-to-end execution, producing concept graphs with sentence-level evidence. This work reframes natural-language formalization as a convergence problem over meaning dynamics and relational self-adjustment.