ad-hoc-semantic-reconciliation

The proposal — Structuring Meaning for Machine Cognition

Ad hoc semantic reconciliation between divergent network models, using machine cognition.

A two-part research proposal with companion materials. It examines how networked systems that reason for themselves need information structured as meaning, not merely form — and how two such systems can reconcile their divergent models ad hoc, with no standard agreed in advance.

The empirical realisation of this programme — four scenarios built, run, and reported — is in ../study/. The interactive demonstrations are in ../demos/.

Papers

Part I — Structuring Meaning for Machine Cognition: Semantic Models, the Cognition Spectrum, and Shared References. Read inline: Part_I_Theory.md · PDF: Part_I_Theory.pdf · slides: Part_I_Theory_Deck.pdf. Covers semantic models as completeness for comprehension, the lift that produces them, semantic-level reconciliation, the cognition spectrum, shared references, and four worked demonstrations.

Part II — A Research Programme on Shared References for Ad Hoc Semantic Reconciliation. Read inline: Part_II_Programme.md · PDF: Part_II_Programme.pdf · slides: Part_II_Programme_Deck.pdf. Sets out the research questions and hypotheses, a two-track methodology, metrics including language-model-based cognitive-effort measurement, a benchmark with datasets, a phased plan, and the contributions sought from operational networks.

(Each paper is provided both as markdown — which renders here on GitHub — and as the original PDF, which is the faithful, printable version. The slide decks are PDF only.)

Interactive demonstrations

Four self-contained HTML studies, one per operational scenario. Clicking an .html in the repo shows its source — GitHub never renders HTML inline — so to see one run, use view in browser (opens it rendered via htmlpreview, no setup) or download it and open the file locally. The source files are in ../demos/.

• Configuration — two peer models of one transport network (TAPI ↔ IETF TEAS); pre-empts false cognates through a thin reference, verified by round-trip on invariants. ▶ view in browser · ⤓ download
• Intent — a declarative, quantitative intent refined to a realisation (TM Forum intent ↔ IETF L1CSM); verification through satisfaction rather than equality. ▶ view in browser · ⤓ download
• Cross-domain — two bespoke systems in different domains with no shared standard; they build a minimal shared reference between themselves. ▶ view in browser · ⤓ download
• Observability — reconciles anomaly semantics across an incumbent fault model and the newer IETF NMOP model, where pragmatics carries most of the meaning. ▶ view in browser · ⤓ download

Suggested reading order

Begin with Part I for the framing and terminology; explore one or two demonstrations to see the method in action; then read Part II for the proposed research programme. From there, the ../study/ reports show the programme carried out and measured.

References

The standards, related community work, and foundational sources supporting the proposal are listed in each paper’s own reference section (Part I, Part II) — spanning IETF RFCs and drafts (NMOP fault/anomaly/incident work, TE topology, L1CSM, the knowledge-graph and NAIM efforts), the ONF Transport API, TM Forum intent models, and the theory of signs and the data-to-wisdom ladder.