ThemeGraph: Semantic Representation Authorization for Directed Text-Analytic Graphs
ThemeGraph is a representation-authorization architecture for human-facing directed text-analytic graphs. At claim time, one policy declared, typed certificate jointly governs the statement, graphical mark, requested-claim export, independently supported fallback, witness, and remediation. Asymmetric normalization of unordered co-occurrence can otherwise yield a numerically correct arrow without evidence of temporal, prerequisite, entailment, or causal direction. Participant-free evaluation compares four generated output-control conditions, removes contract distinctions, tests policy conformance and construct-boundary routes on five datasets, and evaluates a 24-configuration specification surface. All eight currently inventoried certificate-guarded output entry points require valid certificate state. These results establish bounded mechanism behavior, not policy validity, source truth, perceptual effectiveness, usability, analyst benefit, causal truth, or universal non-bypassability.
Authors
- Isaac Oluwafemi Elesemoyo (ORCID: https://orcid.org/0000-0001-9147-3935)
- Yashin Brijmohan (ORCID: https://orcid.org/0000-0003-2485-9808)
- Taiwo Raphael Feyijimi (ORCID: https://orcid.org/0009-0009-5752-4175)
- Clement Feyijimi
Institutions
- Utah State University (US)
- University of Kansas (US)
- Obafemi Awolowo University (NG)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-10-29
- DOI
- https://doi.org/10.5281/zenodo.23756601
- Primary Topic
- Access Control and Trust
- Type
- preprint