GraphNews

Part 2 of from Data Engineering to Knowledge Engineering

The SPARQL query language is the standard method to access knowledge graphs (KGs). However, formulating SPARQL queries is a significant challenge for non-expert users, and remains time-consuming for the experienced ones. Best practices recommend to document KGs with competency questions and example queries to contextualise the knowledge they contain and illustrate their potential applications. In practice, however, this is either not the case or the examples are provided in limited numbers. Large Language Models (LLMs) are being used in conversational agents and are proving to be an attractive solution with a wide range of applications, from simple question-answering about common knowledge to generating code in a targeted programming language. However, training and testing these models to produce high quality SPARQL queries from natural language questions requires substantial datasets of question-query pairs. In this paper, we present Q²Forge that addresses the challenge of generating new competency questions for a KG and corresponding SPARQL queries. It iteratively validates those queries with human feedback and LLM as a judge. Q²Forge is open source, generic, extensible and modular, meaning that the different modules of the application (CQ generation, query generation and query refinement) can be used separately, as an integrated pipeline, or replaced by alternative services. The result is a complete pipeline from competency question formulation to query evaluation, supporting the creation of reference query sets for any target KG.

Combining IT Infrastructure, Service, and Security Operations Data for Real Insights

A pragmatic approach to interoperability: model your data richly for your own needs, then allow the eventual creation of one‑to‑many alignments.

A personal note on AI music fatigue, distrust of rec engines, and using Wikidata queries to browse like a record store.

Learn how JanusGraph with Azure Managed Cassandra enables scalable, high-performance graph applications for complex relationship analysis.

Clustering, Cognition, and Ontology Engineering: The Proven Foundations We’re witnessing a powerful convergence in modern knowledge engineering. At its heart lies a decades-old pairing: the TBox-ABox architecture of description logics. This split between a schema layer (TBox: classes, axioms, rules) and an assertional layer (ABox: instances, facts) is more than a modeling convenience—it is the structural backbone for scalable, interpretable, and efficient knowledge systems.​ 1. Bidirectional TBox-ABox Reasoning: The TBox-ABox relationship isn’t just a one-way street. Modern ontology platforms implement feedback cycles—TBox-driven reasoning enriches ABox data, while patterns discovered in ABox instances suggest new or improved TBox axioms. Automated systems like DL-Learner and AMIE+ exemplify this virtuous loop, blending logical inference with machine-driven discovery.​ 2. Iterative Refinement and Machine Learning Analogies: While not mathematically identical, the iterative cycles in clustering algorithms (such as EM or K-Means) echo the knowledge-refinement process in ontologies. In one, the cluster centroids and assignments are repeatedly updated to improve fit. In the other, TBox rules and ABox facts are successively refined for logical coherence and empirical richness. This analogy, though conceptual, powerfully illustrates the self-improving nature of mature knowledge systems.​ 3. Horn Logic for Computational Efficiency: Horn logic unlocks polynomial-time reasoning, the foundation of OWL 2 RL and similar profiles. By restricting to Horn clauses, systems achieve tractable, scalable inferences that keep production-grade knowledge graphs accurate and responsive—even at enterprise scale.​ 4. DL Safety and Rule Guarantees: Combining rules (SWRL) with expressive ontologies (OWL DL) is notoriously tricky—adding rules risks undecidability. The DL-Safety condition was a breakthrough, enforcing restrictions that ensure automated reasoning stays sound and terminates, without sacrificing the power of rule-based schema enrichment.​ 5. Dual Validation for Trustworthy Knowledge: Ontology-based data access (OBDA) completes the story by empirically grounding logic-driven schemas in live production data, while reasoners like HermiT enforce logical consistency. This dual validation ensures that knowledge systems are both scientifically sound and operationally reliable.​ Key Takeaway: Today’s most robust knowledge systems don’t rely on black-box heuristics or wishful analogies. Their power comes from rigorous, validated architectures: bidirectional ontology refinement, tractable logic fragments, and principled integration of logical and statistical learning. This is the true engine behind scalable, explainable, and trustworthy enterprise AI. #OntologyEngineering #KnowledgeGraphs #MachineLearning #SemanticWeb

Understanding Failure Scenarios from LLM-based Entity & Relationship Extraction

This codelab teaches you how to build intelligent multi-agent systems using Google’s Agent Development Kit (ADK) combined with Neo4j graph databases and the Model Context Protocol (MCP) Toolbox. You’ll learn to create specialized GraphRAG-powered agents that leverage knowledge graphs for context-aware query responses, implement agent orchestration patterns, and deploy pre-validated database queries as reusable tools. By the end, you’ll have built a production-ready investment research system demonstrating best practices for next-generation retrieval agents.

How to easily start constructing knowledge graphs using well known data engineering tools

To all my Information Architect peers, my Ontologist friends, and my fellow Knowledge Engineers: Look away, keep scrolling. This post may…

How RDF, SPARQL, and OWL Transform Graph RAG from Pattern Matching into Intelligent Knowledge Retrieval (github repo included at the end)

I conducted a deep research project on the history of knowledge graphs using Gemini's Deep Research Agent and extracted a knowledge graph from the research report in one pipeline. You can find the code that I used to build the agent inside a Colab notebook labelled "KG_Research_Agent.ipynb" inside this GitHub repository: https://lnkd.in/dXsYb_4V I gave the agent the simple prompt: "Research the history of knowledge graphs", and it came up with a detailed report, along with references, which you can read here: https://lnkd.in/erv-aPqK Connected Data Oxford Semantic Technologies Google DeepMind Gephi GitHub #knowledgegraphs #agents #deepresearch #semanticweb

Table of Contents

Learn graph database fundamentals, query languages, and implementation patterns for AI systems. Covers GraphRAG, knowledge graphs, and performance optimization.

PDF | Perhaps the most fundamental notion underlying the desiderata for a successful Semantic Web is Semantic Interoperability. In this context,... | Find, read and cite all the research you need on ResearchGate

A Response to Critical Discourse on Ontology-Grounded Knowledge Graphs Yesterday I published an article on the engineering realities of building BFO-grounded knowledge graphs at scale. The response from the enterprise architecture community surfaced a critical architectural question that deserves de

Beyond the Hype: A Practitioner's Perspective on Ontology-First Architecture The knowledge graph market has exploded with vendors claiming semantic capabilities, ontological reasoning, and enterprise-grade inference. Yet a significant gap exists between marketing collateral and engineering reality.

The context graph conversation has exploded over the past two weeks. Foundation Capital's "Context Graphs: AI's Trillion-Dollar Opportunity" kicked it off.

Co-authored with Gigi Shannon Every taxonomist knows the pressure of delivering the most accurate models under tight timelines and tighter budgets. Now that AI has permeated enterprise tools and strategic plans, the question isn’t whether they should use LLMs, but how to make them work most effectiv

Why the next wave of enterprise AI isn’t about better models, it’s about capturing the “why” behind every decision.

Adaptive Reasoning Graph (ARG) - A deterministic protocol for agents with evolutive long-term memory

Before LLMs, Palantir was competing with Snowflake and Databricks.