Working Papers

Semantic Knowledge Production

Semantic knowledge is increasingly produced through combinations of human expertise, computational processing, external knowledge resources, and artificial intelligence. This paper develops an architectural account of semantic knowledge production for systems in which assertions may depend on different forms of evidence, competence, inference, and prior semantic states while remaining incomplete, contested, or revisable. The framework distinguishes four semantic roles—Evidence, Knowledge, Provenance, and Normative consequence—and treats knowledge production as a progressive process of observation, interpretation, review, inference, and stabilisation. Stabilised semantic states are operationally reusable without being treated as epistemically final, while provenance preserves the lineage and warrant of knowledge-producing activities. A running cultural-heritage example illustrates how heterogeneous evidence and human and computational activities can participate in this process. A developing companion paper, The Review Algebra, formalises candidate generation, review, and finalisation as provenance-bearing transformations between semantic states.

Download & Cite You can download and cite all versions by using the DOI 10.5281/zenodo.22099067. This DOI represents all versions, and will always resolve to the latest one. Author: Daniel Antal | ORCID: 0000-0001-7513-6760 | Google Scholar: Daniel Antal | Contact: Reprex

Bounded Federation: Reviewable Semantic Exchange between Independently Governed Knowledge Graphs

Knowledge-graph federation is commonly approached through distributed query, entity reconciliation, ontology alignment, or the exchange of assertions between graph-based knowledge systems. In digital humanities and cultural heritage, however, independently governed knowledge graphs frequently represent different institutional, disciplinary, or community perspectives and cannot be assumed to share a complete ontology. Requiring global ontology alignment may therefore impose a stronger condition than is necessary for useful semantic interoperability.

This paper introduces bounded federation as a model of purpose-directed semantic exchange between independently governed knowledge graphs. Rather than attempting to establish complete semantic equivalence between participating systems, bounded federation identifies a delimited domain within which sufficient semantic correspondence can be established for a specified purpose. The approach builds on ontology design patterns and competency questions to define the intended competence of an exchange, and uses an intermediate, representation-neutral tuple structure to expose candidate assertions for review.

The model is demonstrated through Wikibase-to-Wikibase exchange between Wikidata and a specialist research Wikibase. Assertions originating in participating graph representations are projected into a bounded relational review space and, where warranted, made available for destination-governed admission or application. The experiment distinguishes between assertions that can be federated, local semantic commitments that require redefinition, candidate exchanges that should be excluded, and subject–predicate intersections that are not applicable because they do not constitute candidate assertions within the federation boundary. The latter distinction demonstrates that a rectangular computational projection may expose potential semantic intersections beyond the actual claim space and that recognising such limits is itself relevant to interoperability.

The paper argues that the quality of semantic federation should not be evaluated primarily by the completeness of ontology alignment or the quantity of exchanged assertions. Better federation is not necessarily fuller federation. A successful federation should maximise warranted semantic exchange while preserving relevant differences and making its boundaries of interpretation, applicability, provenance, inferential competence, and temporal validity explicit; because federation review can expose semantic differences that inform subsequent local governance, bounded federation is also an ongoing governance process between independently evolving knowledge systems.

Download & Cite You can download and cite all versions by using the DOI 10.5281/zenodo.22091552. This DOI represents all versions, and will always resolve to the latest one. Author: Daniel Antal | ORCID: 0000-0001-7513-6760 | Google Scholar: Daniel Antal | Contact: Reprex

Betwixt Reviewable Relational Projections for Semantic Knowledge Production

Semantic knowledge is commonly represented in structures designed for persistence, exchange, or machine reasoning, while human review and much data-processing software operate more naturally on rectangular data. This paper introduces Betwixt, an R implementation of reviewable relational projections for semantic assertions. Betwixt treats atomic subject–predicate–value assertions as independently identifiable semantic units and projects them into tidy long or wider relational forms appropriate to particular review operations. Dual projections additionally align an assertion with another assertion establishing its relationship to evidentiary material, without combining the two into a compound semantic assertion. The resulting representations provide an intermediate layer between semantic models and human-centred review interfaces. The paper distinguishes projection from review semantics, describes the principal projection classes supported by the Betwixt model, and demonstrates their use in constructing reproducible review workspaces while preserving the identities required for subsequent provenance-aware semantic processing.

Download & Cite You can download and cite all versions by using the DOI 10.5281/zenodo.22091535. This DOI represents all versions, and will always resolve the latest one. Author: Daniel Antal | ORCID: 0000-0001-7513-6760 | Google Scholar: Daniel Antal | Contact: Reprex

The Review Algebra

Semantic assertions produced by human and computational processes may be provisional, contested, corroborated, falsified, deferred, or replaced, while their subsequent use may depend on explicit policies determining what is sufficiently stabilised for a particular purpose. This paper develops the Review Algebra, a formal and computational framework for representing these transformations while preserving the provenance-bearing activities through which semantic assertions are generated, examined, and admitted to subsequent semantic states.

The algebra distinguishes three recurring operations: candidate generation, review, and finalisation. Candidate generation makes semantic assertions available for consideration without implying their acceptance. Review examines candidate assertions and returns semantic values together with explicit statuses describing what happened to the candidates. Finalisation applies a policy determining which reviewed assertions enter a subsequent stabilised semantic state.

Observation and inference are treated as distinct mechanisms of candidate generation rather than additional operators of the algebra. Other mechanisms may include reconciliation, federation, import, computational analysis, or human proposal. Their epistemic differences remain represented through provenance while their outputs participate in the same downstream review structure.

The Review Algebra extends ordinary relational and tidy data operations rather than replacing them. Wide and long review workspaces provide computational and human-facing projections over persistent assertions and activities without becoming the canonical knowledge model. Stabilised assertions can subsequently participate in further candidate generation, making semantic stabilisation a recursive, provenance-bearing transformation between reviewable semantic states.

The Review Algebra provides the underlying formalisation for a broader framework of human-centred semantic knowledge production. It deliberately does not prescribe the epistemology of observation, interpretation, reconciliation, or specialist judgement. Instead, it provides a common computational structure through which assertions produced by heterogeneous human and computational activities can remain reviewable, provenance-bearing, and reusable.

Download & Cite You can download and cite all versions by using the DOI 10.5281/zenodo.22091483. This DOI represents all versions, and will always the latest one. Author: Daniel Antal | ORCID: 0000-0001-7513-6760 | Google Scholar: Daniel Antal | Contact: Reprex