Are graphs really the new star schema? What do graphs look like to non-insiders, and what is it that attracts them to the graph community, methodologies, applications, and innovation? Phakathi nenkqubo yokuxhumana Ngaphezu kwalokho, sinikezela ukuxhumana okuhlobisa kanye nokuhlobisa izakhiwo kanye nokuthuthukiswa kwe-community kanye nemikhiqizo esebenzayo: Knowledge Graphs, Graph Analytics, AI, Data Science, Databases, Semantic Technology. I-Connected Data ye-London 2025 Ukulandelana nathi - isisindo se-adopters ezivela eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside eside. Namhlanje, kunezinto ezininzi zebhizinisi zihlanganisa ngaphezulu kunesinye. Kunezinto ezininzi zihlanganiswa ngokukhawuleza kunesinye. Futhi abantu zihlanganisa futhi zihlanganisa ukubaluleka kwezi zihlanganisi ngaphezulu kunesinye. Zonke izinto zihlanganisa. I-graphs iqukethe ukuxhumana kwama-model. Ngakho-ke i-graphs zihlanganisa emhlabeni wonke. Ngenxa yalokho, kukhona isibambo ekubunjweni, umqondo umphathi, izixhobo, izici, izicelo kanye nemikhiqizo ezivela ku-graphs. Lesi sici ye-Year of the Graph ibonise lokhu. Nokho , ne-targeted curation approach akuyona, kuyinto ingxaki elidlulile kuze kube manje. Kuyinto 3 izinyanga ezedlule kusukela ku-previous Okuhlobene wonke - kusuka kumazwe abavela kumadivayisi, futhi kusuka kumadivayisi we-strategic kanye namadivayisi kuya kumadivayisi kanye namadivayisi. Ukusekela i-graph ku-Microsoft Azure, i-GitLab, i-Netflix, i-S&P ne-SAP, kuya ku-semantics ne-agents ku-Databricks ne-Snowflake, i-ontologies, i-knowledge charts ne-AI, ama-graph transformers, ne-science breakthroughs. To meet the leaders and innovators shaping Knowledge Graph, Graph Databases, Semantic Technology and Graph Analytics / Data Science / AI, come to . Connected Data London 2025 on November 20 – 21 Connected Data London 2025 ngoNovemba 20 - 21 Table of Imininingwane Graph is the new star schema Graphs power Systems of Intelligence A unified semantic knowledge graph for Enterprise AGI Defining and building ontologies Getting started with knowledge graphs Adopting, building and populating knowledge graphs Knowledge graphs and AI: a two-way street The state of GraphRAG Multimodal graphs Graph databases grow and evolve LPG vs. RDF, OWL vs. SHACL Graph AI: GNNs, graph transformers and foundational models Graph science: Strong perfect graphs, the new Dijkstra’s algorithm and convergent neural networks Umthombo we-Year of the Graph iyatholakala kwami ngu-G.V(), i-metaphacts, i-Lincurious, ne-cognee. Uma ufuna ukwesekwa ku-edithi elandelayo futhi ukweseka lokhu umsebenzi, ukufinyelela! This issue of the Year of the Graph is brought to you by , , , and . G.V() metaphacts Linkurious cognee Waze () Ukuhlobisa ikhaya ikhaya Uma ufuna ukwesekwa ku-edithi elandelayo futhi ukweseka lokhu umsebenzi, ukufinyelela! Announcing the State of the Graph project Ukuhlaziywa kwe-State of the Graph Project I-repository ephelele futhi ephakeme, ukubonisa kanye nokuhlola zonke izinzuzo ezivela ku-graph technology space. 👉 Kuqala ngokushesha. Register for updates lapha Graph is the new star schema I-Graph is the new star schema I-Irina Malkova, i-VP Product Data & I-AI ku-Salesforce, usho. Ngaphambi kwe-AI, i-Malkova awukwazi ukuthi i-graph metadata iyinhlangano ye-ROI ye-team yayo - kodwa akuyona kakhulu. I-graph iyona ngokwenene i-stars scheme entsha? Ngiyazi, kunjalo, Malkova ibonise ukuthi ama-agents akuyona asebenzayo uma idatha asekelwe njenge-graph. Amahora amabili ngemuva kwe-Malkova wabhala lokhu, umbhalo wama-agents wama-graph-leaders abahlala ukuthi abahlala emkhakheni kanye ne-context. Njengoba wahlala, Malkova ukhangela ukuthi kungcono ukuthi ama-graph insiders kanye nama-newcomers bonke babonisa umsebenzi zabo ngokufunda ngamunye. thina siphinde nge-third wave ye-graphs, eyenziwe ngoku ku-need to feed data to AI agents Yini i-graphs ne-data warehouse star schema zihlanganisa? Lezi zihlanganisa, futhi zihlanganisa ukunambitheka ama-value kumazwe. Ngokungafani ne-star schema, kunjalo, amamodeli we-graph data zihlanganisa futhi angakwazi ukucubungula semantics ngokunambitheka. Thina indlela eyodwa - i-graph schema iyona i-star ye-schemes. Ukulungiswa Ngaphandle kwezinkqubo ezivamile nezakhiwo ezinzima, isivinini esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe esithathwe. I-Charles Betz, VP, Umphathi we-Chief Analyst ku-Forrester Ukulungiselela ukwakha imodeli esebenzayo ye-IT esebenzayo esihlanganisa i-GenAI ku-structure ye-management systems. I-Graph databases kanye ne-recovery-augmented generation zihlanganisa ubuchwepheshe yayo. I-Graphs inikeza izakhiwo kanye nezinhlangano ngokufanelekileyo, okuvumela i-GenAI ukubuyekeza ngezindawo ze-data emangalisayo. Ngokuvimbela ku-graph-based knowledge infrastructure, izinhlangano angakwazi ukuvikela umthamo ephelele we-AI nangokuphepha ukubuyekeza, ukucubungula, kanye nokuhlanganiswa. U-Matan-Paul Shetrit, umphathi we-Product Management ku-Writer, uye uye uxhumane okuhle, Ngama-Enterprise ezivamile, ukuhlanganiswa kuyinto manual. Ngama-Enterprise e-hybrid, ukuhlanganiswa kwenziwa ku-programable. Ukubonisa i-graphs njenge-orchestration layer ye-firms ye-future I-Orchestration Graph kuyinto: inethiwekhi enhle, okungenani engaziwa yabasebenzi, ama-agents, kanye nezinhlelo ezihambisana nge-delegation logic, izindlu zokusebenza, nezindlela zokuphumula. I-G.V() kuyinto i-graph database client ne-IDE enikezela kuzo zonke izivakashi: No matter what graph technology you work with, G.V() makes you more productive Ukubhalisa, ukuqhuma, kanye ne-profile requests Ukuhlola imodeli yakho yedatha Ukuhlola idatha akho nge-graph visualization ephezulu Ukongeza noma ukuguqulwa idatha ku fly Njengoba i-IDE ye-graph database ye-compatible kakhulu, i-G.V() inikeza izixhobo ezingaphezu kuka-20, kuhlanganise ne-Amazon Neptune, i-Google Spanner Graph, i-Neo4j, ne-JanusGraph - futhi manje nge-GQL support ye-Ultipa Graph. Thola wena futhi uqala ukuchofoza database yakho ngaphansi kuka-5 imizuzu: ikhaya Graphs power Systems of Intelligence I-Graphs Power Systems ye-Intelligence I-Graphs ikakhulukazi nge-GenAI, okuphindwe kanye nokuphindwe yi-GenAI. I-Generative AI capabilities ihamba ngesivinini eside futhi izixhobo ezizayo ziye ziye ziye zitholakala ngexesha elilandelayo ze-2-5 ziye ziye zihlanganisa. Ngokusho Gartner 2025 AI Hype Cycle Uluhlu we-Al-Investment ivame kakhulu, kodwa isikhokelo ivela ukusuka ku-GenAl hype ku-innovations eziyinhloko ezifana ne-Al-ready data, i-Al agents, i-Al engineering ne-ModelOps. Ukukhula okusheshayo kwezi zenzakalo nezinkqubo zinikezwa ngokushesha, njenge-hype efanelekayo, okwenza lokhu indawo enhle kakhulu yokuhamba. Lezi zimo zihlanganisa ukuthi i-GenAI iyatholakala ku-priority ephakeme ye-C-suite. I-Knowledge Graphs iyinhlangano ebalulekile yayo, ebonakalayo emkhakheni wokugcwalisa. . I-Knowledge Graphs kuyinto isakhiwo seSystems of Intelligence is a term coined by U-SiliconANGLE & theCUBE analysts uDavid Vellante noGeorge Gilbert . Izinsizakalo U-Geoffrey Moore Ukubuyekezwa kanjani i-Snowflake isebenza ngezimo ezintsha zokusebenza I-Systems of Intelligence iyinhlangano yesakhiwo se-entrepreneurship esidumile ngoba ama-AI ama-agents akuyona kuphela njengezimo se-business eyenziwa ku-knowledge chart. Uma i-plattform ivimbela le chart, ivimbela i-default policymaker ye-”why is this happening, what’s coming next, and what should we do?” A unified semantic knowledge graph for Enterprise AGI I-Unified Semantic Knowledge Graph for Enterprise AGI Vellante noGilbert futhi zoom ku-importance ye-semantics ekuthuthukiseni ukuthi Ukuphathelene ne-strategy kunezinto ukuthi izindlela ze-Intelligence, i-Systems of Engagement, ne-Systems of Agency zihlanganisa indlela yokwenza i-flywheel: I-Data Intelligence Playbook ye-Ali Ghodsi ye-DataBricks ivimbela idatha ku-agency advantage I-User Intent Feeds Semantics: Isikhathi esisodwa esifundisa i-catalogue nge-context enhle kakhulu, ukwengeza umqondo idatha ukuze abanye angakwazi ukufinyelela okungaphezu kwegama. I-Semantics Feed Agents: I-Agents ye-arms eyenziwe ngama-context yokwenza imiphumela emihle futhi ekupheleni kokuphendula ngokuvamile. Agents Create Iziphumo: Agents ukunikela imiphumela ngempumelelo ngokuvumelana nezidingo zebhizinisi. I-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data Intelligence ye-Data. U-Ali Ghodsi uye wabelane imikhosi, ukhangela ukuhweba kwe-semantics "i-existential" Umphakeli uyenza ama-agents abaziwa kuphela okufanayo, kodwa futhi angakwazi ukuhlaziywa kanjani, ukuhlaziywa yini elilandelayo futhi ukuhlaziywa kanjani. Lokhu kuyinto i-Sacred Grail ye-Enterprise AI – isisekelo se-“Enterprise AGI.” Ukufinyelela khona kunama-abstract ngaphandle kwe-RDBMS ne-tables, kuya ku-diagram ye-semantic knowledge. Umfundisi we-strategic akufanele ukwahlukanisa isisekelo se-semantics, futhi Kuyinto ingxenye yezizathu ukuthi abantu zihlanganisa izindandatho semantic - kodwa yini isindandatho semantic, ngokwenene? izindwangu semantic zihlanganisa kakhulu kunoma kwe-AI Umphakathi we-Connected Data u-Sofus Macskássy, uJessica Talisman, uJuan Sequeda futhi uAndreas Blumauer . izigaba semantic ukusuka amazinga eziningi, ukuxhumana izici kanye izicelo Why AI alone can’t solve all your data problems I-LLM ikhona "i-hallucinate" ngoba zihlanganisa kuma-patterns amakhulu, amasethi, futhi akuyona ulwazi olulodwa kwebhizinisi yakho. Ukukhishwa kwe-AI kumamodeli yakho ye-semantic ivimbele ukukhipha ngokunembile futhi enokutholakalayo ngokuvimbela ukuxhumana kwezobuchwepheshe ngaphakathi kwedatha ye-enterprise. I-metis iyinhlangano lwe-intelligence eyenziwe nge-intelligence eyenza idatha eyenziwe ku-business value. Ngokusebenzisa i-metis, ungakwazi ukwakha futhi ukulawula amamodeli we-semantic nge-AI, ukwakha kanye nokuthumela ama-agent eyenziwe ngqo, kanye nokuxhumana izixhobo zokufaka, ukuguqulwa kwebhizinisi nokunye. Ukusebenzisa i-AI enokutholakalayo, enokutholakalayo enokutholakalayo ibhizinisi lakho. Ukuhlola i-metis namhlanje! Defining and building ontologies Ukubonisa futhi ukwakha ontologies Ukuqhathanisa izicathulo semantic kungenziwa meta-semantics, okuyinto i-ironic, kodwa ngokulandelana nezidingo. Ukuqhathaniswa okucacileyo kuyinto semantics kuyinto, futhi nge-interest ezintsha kanye nezivakashi ezivela ku-field, ukuxuba, i-hype ne-arguments okushisayo zihlanganisa imiphumela emibi. Sama kungenziwa malunga ne-ontology. Uma i-CTO ibonise " ” futhi abacwaningi zihlanganisa, umxokozelo kungatholakala ngokushesha ngokushesha. Olandelayo zihlanganisa izindawo ezingaphezu zokusebenza, ezifana . Yini i-ontology What organizations can learn from the poster child for ontology in use: Palantir Ngokuvamile, kungcono ukuthi umzimba wahlukanisa umzimba we-intestinal. Ngathi ? I-ontology engineering ngokwenene ingxoxo Why are ontology and data architecture teams solving the same problems with different languages I-Palantir ine-definition yayo yayo yayo yayo yama-ontology yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo yayo. , futhi i-group thinks i-limits evolution. izinga ezivamile zihlanganisa izicelo I-Consensus yokukhula ukuthi U-Joe Hoeller isibopho sokuthuthukiswa kwe-ontology kubalulekile ngenxa yokuxhumana nabasebenzi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi abacwaningi. I-ontologies isithombe, okungenani kunokwenzeka ukuvelisa I-Top-Level Ontologies (TLOs) njenge-Basic Formal Ontology (BFO), i-Common Core Ontology (CCO), noma i-SUMO (I-Suggested Upper Merged Ontology) zihlanganisa izincwadi ezinzima. . I-TLOs inikeza u-ROI ngezindlela ezimbili Okokuqala, ukuhlangabezana nokuvumelana – akufanele ukubuyekeza noma ukuguqulwa kwezigaba eziyinhloko kuzo zonke iiprojekthi. Okokuqala, lapho amazwe kufuneka ukuhlangabezana – yokufinyelela nge-logistics, ukuhlangabezana nezinsizakalo, ukwelashwa nge-insurance – isakhiwo esihlalweni ukunciphisa kakhulu izindleko zokuxhumana. Kuyinto okuyinto abantu ezifana neTony Seale zitholela. I-Stardog CEO ne-Founder, i-Kendall Clark ibonisa ukuthi . Ukusetshenziswa kwe-LLM ukweseka ukuthuthukiswa kwe-ontology building ontologies leveraging foundational models is “magic” that works via prompt scaffolding, symbolic alignment, formal encoding and iterative validation Nokho, Kuyinto, wathi, kusekelwe ku-assumption ukuthi kufuneka ukwakha i-ontologies eziningi, njengoba yonke inkqubo kuyimfuneko i-ontology. U-CEO & Umbhali we-Semantic Arts uDave McComb utshintshe umqondo ukuthi ukufumana i-LLM yokwenza isakhiwo se-ontology enhle kuyindlela elungileyo yokuqala Yini ufuna, McComb arguthile, kukhona orders ububanzi amancane ontologies. Uyakwazi ukubona superpowers of ontologies lapho unayo ingcindezi elula kakhulu kunoma izakhi zokusebenza. Lezi zimo zihlanganisa I-Open-Source, i-ontology ye-business-focused eyenziwe ngempumelelo yi-Semantic Arts. Isakhiwo se-Lightweight kanye nokusetshenziswa kwe-Current Terminology kunikeza ithebula elula yokukhuthaza ukuthuthukiswa kwe-ontology ye-domain ehlanganisiwe. . ikhaya gist iyahambisana BFO Ngaphandle kokufaka kanjani ukubuyekeza i-ontology yakho, izixhobo kanye nemikhiqizo zokuxhumana ziye zitholakala. , izixhobo ezifana ne open source Waze . Ukuhlolwa kwe-ontology engineering I-Spreadsheet-based Ontology Maker I-OntoAligner, i-Modular and Robust Python Toolkit ye-ontology alignment Sneak peek: Graph visualization and analytics, reimagined for the cloud Ukwakhiwa ngo-2013, i-Linkurious inikeza amabhizinisi ze-2000 ze-Global kanye nezinhlangano zomthetho ukuguqulwa idatha ehlanganisiwe kwama-insights eqinile. U-Linkurious Enterprise Cloud (izinsuku ezingenalutho kuphela ukusungulwa ... uye phakathi kwama-first ezivela!) iyindlela enhle kakhulu yokusebenzisa, enhle futhi enhle yokufaka idatha yakho ye-graph. Ngezinye imizuzu, ukwakha i-akhawunti, ukuxhuma idatha yakho ye-graph, futhi ukuhlola izilinganiso ngokubanzi - akukho isakhiwo noma ukugcinwa okwenziwe. Get early access – join the waitlist for a free trial. Getting started with knowledge graphs Ukuqala nge-Knowledge Graphs I-layer ye-semantic eyenziwe nge-knowledge charts ikhiqiza i-value eyenziwe yi-data ngokuvumela ukuhlanganiswa kwe-data ngokushesha futhi ukwandisa ikhwalithi ye-data kanye nokuphuculwa kwama-contextual relationships kanye nemodeli ezahlukahlukene. Ukubuyekezwa kwe-SUMIT PAL Izinhlelo ezintsha ze-ECL (Extract, Contextualize, Load) kunezinto ezivamile ze-ETL (Extract, Transform, Load) ukuze zithuthukisa ROI ku-AI Okokuqala, ngoba idatha yakho ye-tabular kubaluleke, ngisho nangokuthi ibhizinisi yakho ayikho. Futhi okwesibini, ngoba ngaphandle kokuqinisekisa kwe-predictive (i-statistical) esekelwe ku-LLMs, ukufakelwa kwe-ontological (i-logical) esekelwe ku-knowledge charts kuyinto yokuqinisekisa nokucaciswa. Izinkampani zihlanganisa AI yabo ku-knowledge graphs Ungathanda ukuthi i-knowledge charts iyinto emangalisayo kakhulu yokusebenza, noma ukuthi zihlanganisa ama-datasets amancane. Lezi zihlanganisa kuphela kwezinye izicelo ezivamile zokusebenza. U-Vasilije Markovic uye uxhumane. Kodwa-ke uma unemibuzo, kunezinto ezininzi ukuze uqala. Izinzuzo ezizayo ezivamile ezivela ku-knowledge graphs Ngena » “, uFrank Blau inikeza umhlahlandlela we Graph Thinking. Ku “ I-series, i-Paco Nathan ibonisa indlela yokufinyelela ngesivinini ku-graph fundamentals. I-Max De Marzi ibhekisela . And . Ukuhlobisa Graph Graph ubuchwepheshe Demystified Graph Ukuhlobisa Mastery Tips Uma ufuna ukufinyelela umkhakha we-Knowledge Graph Engineer, uThomas Thelen wahlala idivayisi lokuphendula Ukwakhiwa kwe-knowledge chart kuso enhle, kodwa ukufinyelela ngokushesha ngaphandle kokuphendula kwezinye imibuzo eziyinhloko zokusebenza kunokukhathazeka isikhathi kanye nokushintshwa kwe-over-engineering. Sabika Tasneem isihlalo . 15 questions to help you start smart, whether you’re building a simple internal graph or planning a complex GenAI-powered system Ukukhiqizwa kwe-enterprise knowledge graph is a multi-phase journey. Njengoba amaphrojekthi ukuguqulwa kusuka ku-proof-of-concept yokuqala kuya ku-producalized, multi-domain graph, izindleko zihlanganisa. . U-Joe Hoeller iveza ngezinyathelo ezivamile (PoC, i-pilot, ne-enterprise deployment ephelele) kanye ne-cost eyenziwe ngamunye Cognee turns any data into a queryable knowledge graph backed by embeddings Cognee ukuguqulwa idatha unstructured, eyakhiwo futhi semi-structured ku i-cheryable knowledge graph ehlanganisiwe nge-embeddings. Cognee retrievers blend vector similarity with graph traversal for precise, multi-hop answers and reproducible context – so agents reason and remember with structure. Add Cognee’s enrichment layer, time-awareness, auto-optimization, and its new UI for an even better experience. For teams building domain-aware agents, copilots, and search for knowledge-heavy domains. Cognee is open source, with a hosted version – cogwit. Try it. Cognee is open source, with a hosted version – cogwit. Try it. Adopting, building and populating knowledge graphs Adopting, building and populating knowledge graphs I-Knowledge Graph Ukuvumelana kuyinto enhle. Kukho abantu abasebenzisi abasiza ukwakha i-Knowledge Graphs, kanye nezindlela ezininzi zokwenza lokhu kunoma. Ngaphandle kwe-SAP, ungakwazi ngoku Ngathi ingasetshenziselwa i-codebase RAG, i-code navigation, i-impact analysis kanye ne-architecture visualization. . use semantic querying with the SAP HANA Cloud knowledge graph I-GitLab ye-Knowledge Graph I-Graph-Code iyinkqubo ye-Graph-based ye-open source ye-RAG ye-codebase Synalinks is a Keras-based neuro-symbolic framework that bridges the gap between neural networks and symbolic reasoning. Iziqu ze-optimized ne-constrained knowledge graph extraction and retrieval, ukuhlanganiswa ne-agents, ne-Neo4j support, i-Cypher query generation kanye ne-automatic entity alignment. SynaLinks latest release 0.3 I-Cognee iyinkqubo ye-modular ye-end-to-end knowledge graph construction and retrieval. I-post ye-joint by the cognee ne-Kuzu teams ibonisa . Ngaphezu kwalokho, Amber Lennox akhawunti . how to transform relational data into a knowledge graph how to go from raw data to a knowledge graph with SynaLinks There’s no shortage of tutorials on other tools either. Gal Shubeli shows . Thu Hien Vu shares , and Alain Airom . how to build a knowledge graph from structured & unstructured data using FalkorDB and Graphiti how to extract knowledge graphs from text with GPT4o builds a knowledge graph from documents using Docling (Ukuhlukaniswa kwe-semantic triples) kusuka kumadokhumenti usebenzisa indlela ye-agentic, i-ontology-driven. It combines management ye-ontology, processing ye-language natural, kanye ne-knowledge graph serialization ukuze ukuguqulwa kwe-text ebonakalayo ku-data ebonakalayo ebonakalayo. ikhaya I-OntoCast iyinkqubo ye-open source yokwenza i-knowledge graphs is a framework for fully autonomous knowledge graph construction that eliminates the need for predefined schemas. It leverages LLMs to extract knowledge triples and induce comprehensive schemas directly from text. . AutoSchemaKG AutoSchemaKG is cutting edge research, with the code released on GitHub markets a system that turns PDFs and text files into knowledge graphs. . Blue Morpho iText2KG, an open source Python package designed to incrementally construct consistent knowledge graphs with resolved entities and relations, can now build dynamic knowledge graphs , which uses LLMs to automate the most painful part of building knowledge graphs from text: deduplicating records. I-Russel Jurney ibonise indawo ebandayo ye-semantic entity resolution ye-knowledge graphs Andrea Volpini shares a notebook , futhi Prashanth Rao inikeza . Ukuhlola ukubuyekezwa kwe-entity semantic & ukusetshenziswa kwe-DSPy ne-Google's entsha LangExtract library a gentle introduction to DSPy for graph data enrichment Thola ku-Year of the Graph Newsletter Ukucubungula zonke izinto Graph Year-on-Year Subscribe to the Year of the Graph Newsletter Thola ku-Year of the Graph Newsletter Ukucubungula zonke izinto Graph Year-on-Year Knowledge graphs and AI: a two-way street Knowledge graphs and AI: a two-way street Kwangathi emhlabeni real, . Ukusetshenziswa kwe-entities kanye nama-relationships e-pre-defined kanye nokucubungula ama-duplicates kanye nokucubungula ama-sources e-inconsistency kuyinto impendulo yokwakha ama-knowledge charts. As Panos Alexopoulos notes, these are the types of . And . knowledge graph quality issues that hamper downstream applications trying to automate knowledge graphs may also end up having unforeseen consequences The authors of “ Uyakholelwa ukuthi i-knowledge charts kanye ne-LLM zokusebenza ngokubambisana. Zibonisa indlela yokuhlanganisa i-knowledge charts ngokuvumelana nezidingo zebhizinisi nezithombe zebhizinisi, indlela yokusebenzisa i-ontologies, i-taxonomies, i-data eyenziwe, ama-algorithms ye-machine learning kanye nokuhlanganisa. Knowledge Graphs and LLMs in Action Ngokuhambisana ne-interpretability research, i-Anthropic yasungulwa indlela entsha yokuhambisana ne-"thoughts" ye-model enkulu ye-language. Isisombululo kuyinto ukukhiqiza i-attribution charts, okuyinto (ngxenye) ibonisa iminyango e-model eyenziwe ngempumelelo ku-output eyodwa. , releasing a frontend to explore graphs. Michael Hunger wrote a , futhi Srijan Shukla . I-Anthropic open-sourced library enikezela ukukhiqizwa kwe-attribution graphs ku-open-weights amamodeli ezidumile script to import the graph json into Neo4j open sourced code to transform Claude’s hidden memory into interactive knowledge graphs In the world of LLMs, the term “context engineering” has been getting traction. njengoba "ukwakha izinhlelo dynamic ukunikela ulwazi olufanelekayo nezinsizakalo ngokufanelekayo ukuze LLM kungenzeka ukufinyelela umsebenzi". LangChain’s CEO Harrison Chase defines context engineering As Jérémy Ravenel notes, Futhi uma i-AI iyahamba ngaphandle kwe-demo kanye ne-copy-copies ku-systems eyenza, ukucubungula isithombe, nokuxhumana phakathi kwedolobha, then context alone is not enough. We need ontology engineering. context without structure is narrative, not knowledge Context engineering is about curating inputs: prompts, memory, user instructions, embeddings. It’s the art of framing. Ontology engineering is about modeling the world: defining entities, relations, axioms, and constraints that make reasoning possible. Context guides attention. Ontology shapes understanding. . Knowledge graphs excel at providing structured, semantic context to LLMs by organizing information as interconnected entities and relationships, making them great options for Memory and Retrieval, as . I-Graphs ye-Knowledge eyenziwe nge-ontologies iyinhlangano lokugqibela le-context ye-LLMs Anthony Alcaraz Izindaba Agentic knowledge graph construction and temporal graphs Agentic knowledge graph construction and temporal graphs is the latest in automated knowledge graphs. The idea Anthony Alcaraz promotes based on a tutorial by Andrew Ng and Andreas Kolleger is to deploy an AI agent workforce, treat AI as a designer, not just a doer, and use a 3-part graph architecture to augment humans instead of replacing them. Agentic knowledge graph construction Umbhali we-book » ” aim to equip data scientists to build intelligent AI agents that reason, retrieve, and interact dynamically, empowering them to deploy AI solutions. They dedicate a chapter to creating and connecting a knowledge graph to an AI Agent. Ukwakhiwa kwe-AI Agents nge-LLMs, i-RAG, ne-Knowledge Graphs I-Google Cloud yasungulwa yi-Google Agentspace, okuyinto inikeza isikhungo esisodwa yokwakha, ukulawula kanye nokuthuthukiswa kwamakhasimende we-AI ku-scale kumadivayisi, amaqembu namabhizinisi. . Google Agentspace is powered by a knowledge graph, built on Spanner Graph , claiming to be the largest open source financial knowledge graph built from unstructured data. FinReflectKG is a concrete example of agentic construction and evaluation of financial knowledge graphs and performing multi-hop retrieval directly over those graphs. While the cookbook focuses on OpenAI models and some other specific tooling, the underlying framework and logic are model-agnostic and easily adaptable to other stacks. OpenAI released a hands-on guide for building Temporal Agents with knowledge graphs Uhlobo olulodwa uFareed Khan ku “ I-Khan ibonisa indlela yokwenza ipayipi ye-end-to-end ye-agent ye-time eyenza idatha ashisayo ku-knowledge base enhle, futhi ukwakha uhlelo lwe-multi-agent yokubala ukusebenza. Ukwakhiwa kwe-I.I. Agent ye-Temporary to Optimize Evolving Knowledge Bases e-Modern RAG Systems . TGM is a research open source library designed to accelerate training workloads over dynamic graphs and facilitate prototyping of temporal graph learning methods. It natively supports both discrete and continuous-time graphs. I-Temporal Graph Modeling kuyinto lapho i-TGM isekelwe The state of GraphRAG The state of GraphRAG And what about GraphRAG? Just over a year ago, GraphRAG was the hottest topic in AI. GraphRAG is an emerging set of techniques to enhance retrieval-augmented generation by integrating knowledge graphs, using their structured nature to provide richer, more nuanced context than standard vector search could offer. Several architectural blueprints for harnessing these graphs to capture the complex relationships between entities were laid out, with the goal of producing more accurate and contextually aware AI-generated responses. Since then, Ben Lorica has been watching for signs of these techniques taking root in practice. Nangona imibuzo yokuthola ngokubanzi kunoma, izicelo zangaphakathi zangaphambili zihlanganisa. Ku-agentic AI systems, i-graph iyahluka kusuka ku-data source elula yokufaka ku-map yokufaka nokuhlanganisa. Ngiyaxolisa The true value of the graph-centric approach becomes clear when applied to agentic AI GraphRAG is still hot. Neo4j published the and Avi Chawla isihlalo a , while a Ngathi claims to decrease RAG retrieval error rate by 67%. Developer’s Guide to Graph RAG I-Essential GraphRAG Umbhalo Umhlahlandlela we-RAG vs Graph RAG isifundo se-empirical isifundo lapho futhi indlela yokusebenzisa i-knowledge charts for RAG Anthropic’s Contextual Retrieval There are more new GraphRAG variants too. takes cues from the brain to improve LLM retrieval. HippoRAG combines GraphRAG with Reinforcement Learning. Graph-R1 introduces a novel distillation framework that transfers RAG capabilities from LLMs to SLMs through evidence-based distillation and Graph-based structuring. DRAG I-HiRAG isebenzisa i-clastering ye-hierarchical ukuxhumana ama-clasters ezahlukile ze-theme ukuze kuthuthukise ukubuyekezwa kwe-global. Andreas Kolleger highlights , futhi Ben Lorica isihlalo . innovative approaches from the GraphRAG Track at AI Engineer World’s Fair 2025 5 breakthroughs you should know about in RAG Reimagined Sergey Vasiliev isixazululo Ukuze ukuhlangabezana okuhle, abacwaningi we-Huawei wabhala . GraphRAG doesn’t lack ideas, but it struggles to scale up a pragmatic case study in balancing scalability with reasoning depth in GraphRAG systems Multimodal graphs I-Multimodal Graphs A topic that’s gaining momentum in GraphRAG and beyond is multi-modality. isistimu ye-all-in-one RAG enikeza i-multimodal knowledge graphs for automatic entity extraction and cross-modal relationship discovery for enhanced understanding. RAG-Anything is a framework designed by David Hughes and Amy Hodler to seamlessly integrate visual and textual data for more comprehensive insights and more accurate responses. It combines embeddings that capture visual and audio semantics, graph-based reasoning and explainable outcomes. Multimodal GraphRAG combines structured knowledge representations with deep learning techniques to handle diverse information sources. ikhaya Multimodal for Knowledge Graphs (MM4KG) In “ "Ukuhlola, abacwaningi akhiphe isakhiwo esiyinhlanganisela ye-multimodal graph data, umsebenzi, kanye ne-model, ukuhlola i-multi-granularity enhle kanye ne-multi-scale characteristics e-multimodal graphs. Towards Multi-modal Graph Large Language Model Graph databases grow and evolve Graph databases grow and evolve as Joe McKendrick writes on ZDNet. Graph databases are projected to have a five-year CAGR of 24% – 26% according to and , respectively. The overall database market will grow 16% annually. I-Graph Databases ihamba, ngenxa ye-AI boom, Gartner Inkampani ye-Business Research As AI and RAG have given a significant boost to both graph and vector databases, people are trying to establish how these two compare, and when and how to use each. U-Andreas Blumauer . : they encode logic, preserve causality, and let you do symbolic + neural hybrid search, Shobhit Tankha chimes in. In André Lindenberg’s words: . compares vector and graph database semantics Graphs don’t just store facts A database tells you what is connected. A knowledge graph tells you why Graph databases are bustling with activity. First, we saw the unveiling of not one, but two new vendors in the last couple of months. , an efficient disk-based graph database for RDF knowledge graphs, is now in open beta. And , a low-latency in-memory graph database engine, is now open for early access. Ukuhlobisa Ukuhlobisa Existing graph database vendors are making progress too. , a new graph architecture that aims to eliminate data silos between transactional and analytical systems. Neo4j went HTAP by launching Infinigraph introduced engine improvements and support for AWS Graviton-based r8g instances. Amazon Neptune 1.4.5 , bringing improvements in developer ease, performance, and cost efficiency. Aerospike Graph Database 3.0 was announced , bringing broad LLM compatibility, MCP support, precision entity linking, native GraphQL support and performance improvements. Graphwise announced the availability of versions 11 and 11.1 of GraphDB , bringing single-file databases, improvements to vector and full-text search indices, and new LLM support. Kuzu v0.11.0 was released . TigerGraph announced a strategic investment from Cuadrilla Capital GQL, the newly-minted graph query language standard, is seeing adoption. , enabling users to run GQL queries on any Fabric Eventhouse or Azure Data Explorer. Shortly after, Ukubuyekezwa integrated with deep search. Microsoft is adding GQL support to KQL graph semantics I-Microsoft Fabric uqala ukunikezelwa kwe-graph analysis ku-Real-Time Intelligence Siren is the first investigative platform to offer GQL graph querying For a guide to designing, querying, and managing graph databases with GQL, check the newly released book . Futhi manje ungakwazi ngqo , thanks to the Ultipa VS Code Extensions. Getting Started with the Graph Query Language (GQL) run GQL queries in VS Code Graph data models: LPG vs. RDF, OWL vs. SHACL Graph data models: LPG vs. RDF, OWL vs. SHACL The LPG vs. RDF debate over graph data models never really goes away. Bryon Jacob explored RDF’s complete stack – (NgoLwesithathu) (triples), (RDFS/OWL), (SPARQL), and . Jacob argues that . identity Isakhiwo semantics Ukubuyekezwa compared it to property graphs major enterprises are discovering they’ve been rebuilding RDF piece by piece and Atanas Kiryakov has a go at good insights in the comment section. In , Enterprise Knowledge share ways to manage and apply a selection of these frameworks to meet enterprise needs. Ora Lassila agrees, debunking urban myths about RDF and explaining how ontologies help GraphRAG; bridging LPG and RDF frameworks In his exploration on I-Kurt Cagle inikeza ukuthi ngokushesha uzothola ukuxhaswa kwe-LPG ne-RDF. I-Cagle inikeza futhi ukuthi ukuxhaswa kwe-SHACL-based ku-event-driven kanye ne-dynamic knowledge charts kuyoba kuhle kakhulu. . the future of knowledge graphs arguing it’s time to rethink Linked Data , enabling to transpile domain models into schema definition languages like GraphQL, Avro, SQL, RDF, and Java while preserving semantics. . Netflix unveiled its UDA (Unified Data Architecture) to model once, represent everywhere S&P launched its new AI-ready Metadata on the S&P Global Marketplace, with RDF under the hood I-OWL vs. SHACL. I-Holger Knublauch, i-Boris Pelakh, i-Pete Rivett ne-Jessica Talisman ibheka lokhu ku U-Michael Iantosca wathi ukuthi . I-Big Semantic Modeling Debate U-OWL kanye ne-SHACL angasetshenziselwa ngesikhathi esisombululo se-AI Agents lapho usebenzisa i-knowledge graph Holger Knublauch usihlanganisa ukubuyekeza . Veronika Heimsbakk’s book “ ” is open for pre-orders. And Kurt Cagle shows , and and . Yini kuya ku-SHACL 1.2 SHACL ngenxa ye-Practicer how to make pizza with AI and SHACL how to use SHACL to validate anything Ukulungiselela I-User Interface Graph AI: GNNs, graph transformers and foundational models I-Graph AI: I-GNNs, i-graph transformers ne-foundation models We have already seen how to approach graph data models and databases coming from the relational world. But “Uma? what is a ‘relational foundation model The GenAI boom has given us powerful language models that can write, summarize and “reason” over vast amounts of text and other types of data. But these models don’t work for high-value predictive tasks like predicting customer churn or detecting fraud from structured, relational data. Kumo’s approach, “relational deep learning,” promises to change that. Kumo’s relational foundation model generalizes the Ukubonisa ngokuzenzakalelayo noma iyiphi i-database ye-relational njenge-diagram eyodwa, eyinkimbinkimbi, futhi ucwaningo ngqo kusuka ku-diagram ye-representation. transformer architecture Kumo looks like the first to productize this. However, people in Waze Kuthuthukisa izindlela ezivamile futhi. The evolution is far from over, and the future of graph AI promises to be even more deeply connected. I-Google Yandex I-Janu Verma ye-Microsoft inikeza ukubuyekezwa kwayo ku-Graph Transformers. I-Connected Data Community iyindawo enhle yokuxhumana .The authors of the have recently added a new chapter on graphs. Jure Leskovec shares . introduction to Graph Learning and GNNs Geometric Deep Learning textbook what every data scientist should know about Graph Transformers and their impact on structured data I-PyG (i-PyTorch Geometric) kuyinto ibhizinisi eyenziwe ku-PyTorch ukuze kusebenze kanye nokuhlolwa kwe-Graph Neural Networks ngokunambitheka kwezicelo eziningana nezinkcukacha zokusekelwe. I-PyG iye yandisa kakhulu ukusuka kwelanga lokuqala, okwakhiwa njenge-framework ephambili ye-Graph Neural Networks. Ukubonisa izinzuzo eziyinhloko zokusebenza kwe-scalability kanye nezidingo zokusebenza ze-real world. PyG 2.0 is a comprehensive update . It is a PyTorch-based framework that provides a flexible and modular architecture for building and training GNN models for anomaly detection. I-GraGOD iyisisombululo esebenzayo yokubonisa anomali ye-time-series ngokusebenzisa ubuchwepheshe ze-GNN Graph science: Strong perfect graphs, the new Dijkstra’s algorithm and convergent neural networks I-Graph Science: I-graphs enhle enhle, i-algorithm entsha ye-Dijkstra ne-convergence ne-neural networks Last but not least, advances on the scientific front for graphs. Starting with a . Ukuphuka kweChudnovsky ku Ukubonisa kanjani imathematics abstract ukwakha izixazululo zangempela. I-profil ye-Maria Chudnovsky, i-"superstar mathematician" owaziwa i-puzzle ye-40 edlule mayelana ne-graphs ephelele I-Strong Perfect Graph Conjecture Ngokuba izixazululo ze-real-world: Uma usebenzisa i-Google Maps ukufumana isitimela se-fastest, emzimbeni, kusebenza ingxenye ye-algorithm ye-Dijkstra. Kuyinto yindlela esivamile yokubala "izindlela ezincinane" kusukela ku-1950. : yini abantu abalandeli ukuthi ayikwazanga ngokwenene ngempumelelo ngempumelelo. Researchers have found a faster way to run Dijkstra’s shortest path algorithm ikhaya Travis Thompson Izikhwama U-Alexander Stage wabhala ukuthi lokhu a . okuvumela kahle indlela imikhiqizo data isakhiwa futhi isetshenziswe great theory milestone, but production routing already “changed the rules” years ago with preprocessing and smart graph engineering Miklós Molnár reports on Szegedy Balázs’ work on Iziphumo zihlanganisa ku-Platonic Representation Hypothesis, ngokuvumelana ne-neural networks zihlanganisa ku-shared statistical model of reality. Futhi uAlberto Gonzalez usihlanganisa isisekelo se-Platonic Representation. . Ukuqeqesha i-neural networks nge-architectures ezifanayo Ukubonisa Graphs Thola ku-Year of the Graph Newsletter Ukucubungula zonke izinto Graph Year-on-Year Subscribe to the Year of the Graph Newsletter Subscribe to the Year of the Graph Newsletter Ukucubungula zonke izinto Graph Year-on-Year
