GraphNews

Since I'm not at #ISWC2025, it's more easy for me to speak up. There are ginormous issues with QLever and the associated Sparqloscope benchmark by Hannah Bast and colleagues. The main results table already shows something that's too good to be true. And while I'm sure that table is technically true, the tacit implication that this table has any bearing on real-world performance, is false. QLever is faster than the state of the art… at COUNTing. That's it. QLever can count faster. The implication is that this would mean QLever can also produce results faster. Yet we have zero reason to assume it can—until there's proof. In the real world, query engines rarely compute all results at once. They stream those results. The Sparqloscope benchmark is designed to trick established query engines into actually producing the result set and counting items. And you know what? Sometimes, the established engines are even faster at that than QLever, which seems to be purposefully designed to count fast. Yes—I'm sure QLever is a fast counter. But what on earth does that have to do with real-world streaming query performance? And did I mention that Virtuoso supports SPARQL UPDATE? How can you tell, just from the table? Well, Virtuoso is faster than QLever for just about anything that doesn't rely on pure counting. QLever does “Regex: prefix” or “Filter: English literals” in the ridiculously fast 0.01s? The only rational explanation is that it has a great structure for specifically this kind of counting (again, not results, just count). But Virtuoso is faster for “strbefore”? Well, there you see the real QLever performance when it cannot just count. And only one of those strategies has impact on the real world. So what if a query engine can count faster than any other to 65,099,859,287 (real result BTW). Call me when you can produce 65,099,859,287 results faster, then we'll have something to talk about. In the first place, it's a major failure of peer review that a benchmark based on COUNT was accepted. And I'd be very happy to be proven wrong: let's release the benchmark results for all engines, but without COUNT this time. Then we'll continue the conversation. https://lnkd.in/eT5XrR2k | 19 comments on LinkedIn

Looking to run local GraphRAG or other graph analytics use cases? With DGX Spark, you can prepare your local text files for graph use cases at your desk.In t...

🚀 LLMs can be powerful tools to extract information from texts and automatically populate Knowledge Graphs guided by ontologies given as inputs. BUT how good are they? To reply to this question, we need benchmarks! 💡 With Lettria, we build the Text2KGBench-LettrIA benchmark covering 19 different ontologies in various domains (company, film, food, politician, sports, monument, etc.) and consisting of near 5k sentences strictly annotated with triples conforming to these ontologies (208 classes, 426 properties) yielding more than 17k triples. What's more? We throw a lot of compute to compare the performance and efficiency of numerous Closed LLMs models and variants (GPT4, Claude 3, Gemini) and numerous fine-tuned Open Weights models (Mistral 3, Qwen 3, Gemma 3, Phi 4). ✨Key take-away: when being provided with high quality data, fine-tuned open models largely outperform larger, proprietary counterparts! 📄 Curious about the detailed results? Read our paper at https://lnkd.in/e-EZCjWm See our presentation at https://lnkd.in/eEdCCpdA that I have just presented at the Knowledge Base Construction from Pre-Trained Language Models Workshop colocated with the ISWC - International Semantic Web Conference. You want to use these results in your operations? Sign-up for using the newly released PERSEUS model, https://lnkd.in/e7exyJHc Joint work with Julien PLU, Oscar Moreno Escobar, Edouard Trouillez, Axelle Gapin, Pasquale Lisena, Thibault Ehrhart #iswc2025 #LLMs #KnowledgeGraphs #NLP #Research EURECOM, Charles Borderie

Clinical Knowledge Graph (CKG) is a platform with twofold objective: 1) build a graph database with experimental data and data imported from diverse biomedical databases 2) automate knowledge disco...

🎧 Exciting news! The audiobook version of "Knowledge Graphs and LLMs in Action" is now available! Are you busy but would love to learn how to build powerful and explainable AI solutions? No problem! Manning has just released the audio version of our book. Now you can listen while you're: - Running and training for your next marathon 🏃 - Commuting to the office 🚗 - Sitting in the parking lot waiting for your kids to finish their violin lesson 🎻 Your schedule is packed, but that shouldn't stop you from mastering these powerful AI techniques. Get your copy here: https://hubs.la/Q03MVhhk0 And don't forget to use discount code: lagraphs40 for 40% off! Clever solutions for smart people.

Ontology Bill of Material? Do we really need it? In software engineering, we have SBOMs, Maven, Gradle, pip, and npm. We have decades of best practices for dependency management, version pinning, and granular control. We can exclude transitive dependencies we don't want. In ontology engineering and semantic modeling... we have owl:imports. We're trying to build mission-critical, enterprise-scale knowledge graphs, but our core dependency mechanism often feels like a step back in time. We talk about logical rigor, but we're living in "dependency hell." So: "How do you manage different versions of an ontology? How do you go through the complexity of imports? How do you propagate changes?" And the answer right now is: With great difficulty! and a lot of custom workarounds. The owl:imports axiom is a logical "all-or-nothing" merge. It's defined as a transitive closure. This is the direct cause of our most common and painful problems: - The "Diamond Problem": Your ontology imports Model-A (which needs Common-v1) and Model-B (which needs Common-v2). Your tool just pulls in both, creating a logical mess of conflicting axioms. A software build would fail and force you to resolve this. - Model Bloat: You want to use one class from a massive upper ontology (e.g schema .org)? Congratulations, you just imported the entire thing, plus everything it imports. And good luck with that RAM spikes, lags, ... - No Granular Control: This is the big one. In Maven or Gradle, you can exclude a problematic transitive dependency. In OWL, this is simply not possible at the specification level. You get everything. So, yes, we need the concept of an "Ontology Bill of Materials" (OBOM). We need a manifest file that lives with our ontology (and helps us to build it) and provides a reproducible "build." We need our tools (Protege, OWL API, ...) to treat this as a first-class citizen. This manifest would: -List all direct dependencies. -Pin their exact versions (via VersionIRI or even a content hash). -Resolve and list the full transitive dependency graph, so we know exactly what we are loading. -Detects problematic imports, cyclic dependencies, ... The "duct tape" we use today like custom build scripts, manually copy paste of element and so on are just admissions that owl:imports is not enough. It's time to adopt the mature engineering practices that software teams have relied on for decades. So how do you deal with complex ontology/model dependencies? How do you and your teams manage this chaos today? #Ontology #KnowledgeGraph #SemanticWeb #RDF #OWL | 39 comments on LinkedIn

The O-word, “ontology” is here! Traditionally, you couldn’t say the word “ontology” in tech circles without getting a side-eye. Now? Everyone’s suddenly an ontology expert. And honestly… I’m here for it. As someone who’s been deep in this space, this moment is exciting. We’re finally having the right conversations about semantic interoperability and the relationship with Agentic AI. But here’s the thing: before we reinvent the wheel, we need to understand the road already paved. 🧠 Homework if you’re diving into this space (link in comments): 1️⃣ Read the original Semantic Web vision article by Tim Berners‑Lee, James Hendler & Ora Lassila It laid out a future we’re finally ready for. Before you complain that “it’s complicated” or “that never worked and failed”, recall that this was a vision that laid out a roadmap of what was needed. Learn about the W3C standards that have emerged from this vision. Honored that I got to write a book with Ora! 2️⃣ Explore ISWC (International Semantic Web Conference) This scientific community was created to research what would it take to fulfill the Semantic Web vision. It’s the top scientific conference in this space, running for over 20 years. I’m proud to call it my academic home (been attending since 2008). ISWC will take place next week in Nara, Japan and I’m excited to be keynoting the Knowledge Base Construction from Pre-Trained Language Models Workshop and be part of the Panel: Reimagining Knowledge: The Future and Relevance of Symbolic Representation in the Age of LLMs. Take a look at the program and accepted papers if you want to know where the puck is heading! 3️⃣ Learn the history of knowledge graphs. It didn’t start with Google. It’s not just about graph databases. The Semantic Web has been a huge influence, in addition to so many events over 50+ years that have worked to connect data and knowledge at scale. Prof Claudio Gutierrez and I wrote a paper that goes into this history. Why this matters? Because we’re in a moment where many talk about “semantic” and “knowledge”, but often without acknowledging the deep foundations.  AI agents, interoperability, and scalable intelligence depend on these foundations. The tech, standards and tools exist. If you rebuild from scratch, you waste time. But if you stand on these shoulders, you build faster and smarter. Learn about the W3C standards: RDF, OWL, SPARQL, SHACL, SKOS, etc. Take a look at open source projects like Apache Jena, RDFLib, QLever, Protege. If something’s broken, or if you don’t like how it’s done, don’t start from scratch. Improve it. Contribute. Build on what’s already there. So if you’re posting about ontologies or knowledge graphs, please ask yourself: - Did I look at the classical semantic web work (yes, that 2001 article) and the history of knowledge graphs? - Am I building on the shoulders of giants, rather than re‑starting? - If I disagree with a standard/open source project, am I choosing to contribute instead of ignoring it? | 65 comments on LinkedIn

Discover how Senzing enhances data clarity with AI entity resolution, partnering with NIEMOpen to resolve entities effectively.

Decentralized machine learning - where each client keeps its own data locally and uses its own computational resources to collaboratively train a model by exchanging peer-to-peer messages - is...

Azure Cosmos DB is a multi-model database service that supports various data models, including key-value, column-family, document, and…

🚀 From MSFT earnings call yesterday: Cosmos DB business grew 50% YoY in FY26 Q1! Deeply grateful for our customer's trust.❤️ Proud for Cosmos DB team to be recognized in Satya's earnings call yesterday. Also excited for our sister database SQL Hyperscale for landing nearly 75% YoY growth! Hard work leads to great results - congrats teams! Keep at it! 💪 🎉 https://lnkd.in/gGNfJtMb

Architectural Evolution from GraphX for Big Data and AI Applications

The applied ontology degree will prepare students from around the world for work in this rapidly growing interdisciplinary branch of information science.

Explore top LinkedIn content from members on a range of professional topics.

Free diagram visualization extension for Streamlit Apps.

On behalf of the GraphFrames maintainers, I am happy to announce the delivery of a new release. It is a significant improvement! It improves performance and memory management: The new release provides 3-50x faster performance for all algorithms. The x5 performance improvement in Connected Components is especially important, as it allows one to perform graph-based identity resolution much faster with the new GraphFrames. All Pregel-based algorithms, such as Shortest Paths and Label Propagation, received a boost of around 3x. The new release comes with its own internal fork of Apache #Spark GraphX due to its deprecation in upstream Spark. This allows us to improve the performance of GraphX-based Label Propagation by 50x and fix memory leaks. Now, it is usable for graph processing inside Structured Streaming. New algorithms were added: New algorithms for K-core centrality, cycle detection, and maximal independent set were added. All of them are based on advanced scientific papers and operate fully in a distributed manner. New APIs: A new API for computing vertex degrees based on edge types was added. The motifs finding API now supports undirected, bidirectional, and multi-hop patterns. The #PySpark API has all the recent improvements in the Scala Core, so there is feature parity between the core and Python. Documentation improvements: The documentation has been significantly expanded, especially the sections on the arguments and parameters of the algorithms. To simplify the onboarding process for new users, the documentation website now contains an llms.txt file in the root directory. Asking an LLM chatbot or coding assistant about how to use GraphFrames is now more efficient. It is already published in Maven Central and PyPi! Blog-post: https://lnkd.in/dU4kRmSD

Knowledge Graph Anatomy 101: An Exploration

We are thrilled to be the new stewards of Dgraph at Istari Digital! looking forward to building the next data platform for digital engineering.

Learn how TypeDB combines graph semantics with strong schema and reasoning to model complex domains.

𝐑𝐞𝐮𝐬𝐢𝐧𝐠 Ontologies 𝐌𝐚𝐤𝐞𝐬 Your 𝐋𝐢𝐟𝐞 𝐄𝐚𝐬𝐢𝐞𝐫 𝐃𝐚𝐭𝐚 contains tremendous 𝐯𝐚𝐥𝐮𝐞. Unfortunately, it is often only used in a specific application, even though it would be useful in other contexts as well. However, 𝐬𝐡𝐚𝐫𝐢𝐧𝐠 data is 𝐧𝐨𝐭 a 𝐭𝐫𝐢𝐯𝐢𝐚𝐥 task. To share data effectively within an organization, we need to 𝐚𝐥𝐢𝐠𝐧 our data with a 𝐜𝐨𝐦𝐦𝐨𝐧 𝐦𝐨𝐝𝐞𝐥. The first thought that comes to mind when hearing about the concept of shared data models (also known as ontologies) is often to develop a new one from 𝐬𝐜𝐫𝐚𝐭𝐜𝐡 quickly. That allows for a fast start and often a slow, yet inevitable, 𝐜𝐡𝐚𝐨𝐬. Ontologies aim to provide a well-described and carefully disambiguated meaning. They are about finding consensus, which is a process rather than a quick win. In that regard, using standardized ontologies is tremendously helpful. (1.) Because they are the product of a collaborative process of 𝐞𝐱𝐩𝐞𝐫𝐭𝐬, many potential 𝐩𝐢𝐭𝐟𝐚𝐥𝐥𝐬 have already been considered and 𝐞𝐥𝐢𝐦𝐢𝐧𝐚𝐭𝐞𝐝. They are established and well used. (2.) They are often abstract enough to be 𝐚𝐝𝐚𝐩𝐭𝐚𝐛𝐥𝐞 to more specific domains. Reused ontologies are not a dead end. They are a 𝐬𝐭𝐚𝐫𝐭𝐢𝐧𝐠 𝐩𝐨𝐢𝐧𝐭 for making data your own. (3.) [𝘈𝘯𝘥 𝘵𝘩𝘪𝘴 𝘪𝘴 𝘮𝘺 𝘧𝘢𝘷𝘰𝘳𝘪𝘵𝘦:] They are 𝐛𝐚𝐜𝐤𝐞𝐝 𝐛𝐲 one or more established 𝐨𝐫𝐠𝐚𝐧𝐢𝐳𝐚𝐭𝐢𝐨𝐧𝐬. Often, it is so much 𝐞𝐚𝐬𝐢𝐞𝐫 to 𝐜𝐨𝐧𝐯𝐢𝐧𝐜𝐞 people to use the standard pushed by Google or the guys who standardize the internet itself, rather than your own definitions. That does not mean that there is no need to create your own ontologies. However, your use case is likely not as unique as you think. And it might be useful to extend an existing ontology to your needs or use one as a blueprint. Want to hear more about how graphs can solve your data problems? Join our next webinar: https://lnkd.in/e6JgQzhP

Facebook, one of the world's largest social media platforms, fundamentally organizes its billions of users and their interactions as a vast social network. At the heart of this organization lies the concept of a graph—a mathematical structure consisting of nodes (or vertices) connected by edges (or

Finding the Right Level of Meaning

Knowledge Graphs (KGs) have gained attention due to their ability to represent structured and interlinked information. KGs represent knowledge in the form of relations between entities, referred to as...

The paper 'Property Graph Standards: State of the Art and Open Challenges' (VLDB 2025), Haridimos Kondylakis and his colleagues take an in-depth look at the current state of property graph standards, which form the basis of many modern graph databases. While property graphs have become a popular way to show complex, connected data (think nodes and edges with flexible key–value properties), the ecosystem is still divided. Each vendor or tool implements its own version of 'the standard', which makes interoperability, schema definition and query translation difficult. The authors review the major initiatives to standardise property graphs and demonstrate the current situation: efforts from LDBC, GQL and ISO are advancing the field, but challenges remain. The biggest gaps lie in schema constraints, data validation, and cross-system compatibility — all of which are crucial if graph systems are to become integral components of enterprise data architectures. The paper calls for a unified model in which graph structure, constraints, and semantics are shared across tools and databases. This isn't just academic. It's about ensuring that graph data can be trusted. It's also about making sure that it is portable. And that it can be used at scale. In simple terms, property graphs are maturing. The next step is not just to connect data, but to agree on how we define, validate and exchange those connections. Article: https://lnkd.in/eva_xSsT

"Semantic layer” means different things depending on who you ask. In my latest newsletter, published on Medium first this time, I look at the two definitions and how they can work together. Are you using a semantic layer? if so, which type? #SemanticLayer #DataGovernance #AnalyticsEngineering #DataandAI | 25 comments on LinkedIn

Calling all Graph Explorers! 📣 I'm excited to share that open-source Graph Explorer v2.4.0 is now released, and it includes a new SPARQL editor! Release notes: https://lnkd.in/ePhwPQ5W This means that in addition to being a powerful no-code exploration tool, you can now start your visualization and exploration by writing queries directly in SPARQL. (Gremlin & openCypher too for Property Graph workloads). This makes Graph Explorer an ideal companion for Amazon Neptune, as it supports connections via all three query languages, but you can connect to other graph databases that support these languages too. 🔹 Run it anywhere (it's open source): https://lnkd.in/ehbErxMV 🔹 Access through the AWS console in a Neptune graph notebook: https://lnkd.in/gZ7CJT8D Special thanks go to Kris McGinnes for his efforts. #AWS #AmazonNeptune #GraphExplorer #SPARQL #Gremlin #openCypher #KnowledgeGraph #OpenSource #RDF #LPG

A Practical Builders Guide