Knowledge Engineering for the Generative Era: Redefining and Structuring an Emerging Discipline

Knowledge Engineering (KE) — the discipline of designing systems that acquire, represent, verify, and distribute knowledge — was born in 1977, thrived during the expert systems boom, and collapsed with the AI Winter of 1987–1993. For two decades, it survived in disguise within the Semantic Web, ontology languages, and knowledge graphs. Today, large language models (LLMs) have reignited the need for systematic knowledge management: hallucination rates of 50% to 82% have been documented in adversarial clinical evaluations, and knowledge graph augmentation consistently improves LLM accuracy in QA tasks. Yet the discipline remains fragmented: at least four competing terms describe overlapping practices, no university offers a degree in modern KE, and practitioners operate knowledge pipelines without a shared framework.

This paper proposes a redefinition of Knowledge Engineering for the generative era, structured around a seven-layer knowledge cycle (Discover → Validate → Produce → Distribute → Consume → Feedback → Sustain) and a taxonomy of nine specialization areas. The framework emerged from a documented case study: a scientific content pipeline that produced more than 600 videos over five months using 20+ orchestrated services, where all nine proposed areas arose as distinct operational concerns.

We argue that a 12- to 18-month window exists to formally define this emerging discipline before terminology fragments irreversibly or institutional momentum crystallizes around narrower alternatives, based on Context Engineering's adoption speed (five papers in four months, Gartner endorsement by July 2025) and the typical 12- to 24-month timeline for new university program approval.