Scholarlink

<p>Large knowledge graphs (LKGs) have become essential components in search engines, digital twins, biomedical discoveries, and retrieval-augmented generation. As these graphs scale from millions to trillions of edges, they transform ontologies from passive frameworks into an active semantic control plane tasked with managing ingestion, reasoning, and compliance instantaneously. This article introduces the first comprehensive framework that describes the co-evolution of ontologies and LKGs at a Web scale. We (i) delineate LKGs across six interdependent "V" dimensions and classify reasoning complexity from straightforward lookup to neuro-symbolic inference; (ii) examine four distributed storage models, demonstrating how hybrid solutions integrate RDF, LPG, and columnar lakes to achieve sub-second query times for graphs on a petabyte scale; (iii) develop ontology-driven construction pipelines capable of supporting more than <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(10^{6}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mn>10</mn> <mn>6</mn> </msup> </math></EquationSource> </InlineEquation> triples per second while upholding SHACL constraints with a delay of less than one second; (iv) propose structural and semantic quality metrics and connect them to task-oriented benchmarks like LinkBench and LDBC-SNB; and (v) identify emerging trends, including foundational graph models and self-repairing semantic systems. Collectively, these contributions establish a robust engineering framework for the construction, validation, and evolution of knowledge graphs that are vast, real time, and reliable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The co-evolution of ontologies and extensive knowledge graphs on a web scale

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Sami Zghal,
  • Marouen Kachroudi

摘要

Large knowledge graphs (LKGs) have become essential components in search engines, digital twins, biomedical discoveries, and retrieval-augmented generation. As these graphs scale from millions to trillions of edges, they transform ontologies from passive frameworks into an active semantic control plane tasked with managing ingestion, reasoning, and compliance instantaneously. This article introduces the first comprehensive framework that describes the co-evolution of ontologies and LKGs at a Web scale. We (i) delineate LKGs across six interdependent "V" dimensions and classify reasoning complexity from straightforward lookup to neuro-symbolic inference; (ii) examine four distributed storage models, demonstrating how hybrid solutions integrate RDF, LPG, and columnar lakes to achieve sub-second query times for graphs on a petabyte scale; (iii) develop ontology-driven construction pipelines capable of supporting more than \(10^{6}\) 10 6 triples per second while upholding SHACL constraints with a delay of less than one second; (iv) propose structural and semantic quality metrics and connect them to task-oriented benchmarks like LinkBench and LDBC-SNB; and (v) identify emerging trends, including foundational graph models and self-repairing semantic systems. Collectively, these contributions establish a robust engineering framework for the construction, validation, and evolution of knowledge graphs that are vast, real time, and reliable.