Autonomous Knowledge Engineering Platforms Powered by Multi- Agent Artificial Intelligence: A Comprehensive Review and Framework Analysis

Abstract

Knowledge Engineering (KE) has traditionally been a human-intensive, labor-demanding discipline requiring domain experts and knowledge engineers to manually construct ontologies, define explicit semantic rules, extract relational triplets, and curate formal knowledge graphs. With the emergence of Large Language Models (LLMs) and Multi-Agent Systems (MAS), autonomous knowledge engineering platforms have transitioned from theoretical concepts into practical, self-governing computational architectures. This review presents a rigorous, state-of-the-art synthesis of Autonomous Knowledge Engineering Platforms powered by Multi-Agent Artificial Intelligence. We systematically dissect the fundamental architectural building blocks, including specialized agent roles (such as ontology extractors, entity linkers, consistency auditors, and belief revision agents), inter-agent communication protocols, task decomposition strategies, and consensus mechanisms. Furthermore, we examine the integration of Retrieval-Augmented Generation (RAG), neuro-symbolic reasoning, and dynamic knowledge graph evolution. Numerical and comparative benchmarks demonstrate that multi-agent collaborative platforms achieve up to a 97.8% knowledge extraction accuracy while reducing human-in-the-loop validation overhead by over 80% across complex domains including biomedicine, finance, law, and cybersecurity. Finally, critical research gaps regarding semantic hallucination mitigation, agent drift, multi-agent protocol consensus deadlocks, and scalable enterprise deployment are thoroughly analyzed alongside future research trajectories. KEYWORDS: Autonomous Knowledge Engineering; Multi-Agent Systems;Knowledge Graphs; Neuro-Symbolic AI; Large Language Models; SemanticWeb; Ontology Extraction.