Human-Centred Semantic Knowledge Production

Artificial intelligence can now generate descriptions, classifications, translations, summaries, recommendations, and links between people, places, organisations, events, and concepts at a scale that far exceeds the capacity of humans to review them. As a result, the principal challenge of the AI era shifts from producing semantic information to establishing which information can be trusted, documenting the evidence behind that judgement, and accumulating reviewed knowledge that can be safely reused. This challenge extends across digital humanities, scientific research, public administration, sustainability reporting, legal practice, enterprise information systems, and other domains where decisions depend on trustworthy semantic knowledge. This paper introduces a review algebra for human-centred semantic knowledge production. Situated at the intersection of relational algebra, computer science, information science, and the philosophy of scientific inquiry, the framework represents semantic review as a sequence of explicit operations that progressively transform observations into reviewed semantic knowledge while preserving provenance and evidential history. The review algebra extends naturally from review to inference and operational optimisation, enabling reviewed knowledge to generate new semantic assertions, identify unresolved dependencies, prioritise subsequent review activities, and coordinate collaboration between human experts, AI systems, inference engines, and independent organisations. Unlike existing approaches, which primarily represent metadata, linked data, or provenance, the review algebra represents the production process through which semantic knowledge is created, reviewed, accumulated, and reused. Rather than introducing a new knowledge representation language, it provides a lightweight computational model compatible with ordinary tabular data, relational databases, RDF, Wikibase, and existing interoperability standards. The framework is demonstrated through a multilingual cultural heritage case study involving namespace management, language review, rights clearance, and semantic inference. We argue that future interoperability depends not only on exchanging data but on enabling organisations, computational systems, and domain experts to accumulate and reuse reviewed semantic knowledge. By making semantic dependencies explicit, the review algebra reduces duplicated review effort, supports operational optimisation, and provides a foundation for trustworthy human-centred AI.