\textsc{Text2Graph}: Combining Lightweight LLMs and GNNs for Efficient Text Classification in Label-Scarce Scenarios. arxiv.org/abs/2512.10061

Are LLMs good at populating Knowledge Graphs based on a set of text documents? This is the research question Text2KGBench-LettrIA: A Refined Benchmark for Text2Graph Systems set out to address. Recent advances in Large Language Models (LLMs) have catalyzed significant research into automated knowledge graph (KG) construction from text, a fundamental challenge at the intersection of natural language processing and semantic web technologies. However, the reliability of evaluating model performance is hindered by limitations in existing benchmarks like Text2KGBench, which exhibit shortcomings in data quality, ontological consistency, and structural design. To address these issues, new research introduces Text2KGBench-LettrIA, a substantially revised and curated benchmark derived from the DBpedia-WebNLG portion of Text2KGBench. Primary contributions include: 1) Systematic refinement of 19 domain ontologies to enforce hierarchical structure and formal typing 2) Complete re-annotation of 4,860 sentences, yielding over 14,000 high-fidelity triples under a strict set of annotation guidelines; and 3) Introduction of an enriched data format with enhanced metadata to ensure reproducibility and support multifaceted evaluation The utility of the benchmark is demonstrated by evaluating a suite of both proprietary and open-weights LLMs in zero-shot and fine-tuned settings, respectively. Results reveal a key finding: smaller, fine-tuned open-weights models can achieve superior F1 accuracy compared to their larger, proprietary counterparts, underscoring the critical role of high-quality, schema-aligned training data. Paper: ceur-ws.org/Vol-4041/paper3.… Presentation: docs.google.com/presentation… #Research #LLMs #Benchmark #ISWC25 #OpenData #OpenSource -- The Year of the Graph's Autumn 2025 newsletter issue on all things #KnowledgeGraph, #GraphDB, Graph #Analytics / #DataScience / #AI and #SemTech is out. Subscribe and follow to be in the know. Reach out if you'd like to be featured 👇 yearofthegraph.xyz/newslette…

Another day, another Kuzu fork: Introducing RyuGraph and Predictable Labs The news hit the graph community recently: the Kuzu Github repository was archived on October 10, 2025 as the Kuzu team moved on to new pursuits. KuzuDB is an embedded open source graph database that was pretty popular. KuzuDB is designed for analytics on large databases and includes full text search, vector indexes, query parallelism for fast performance on multi-core hosts, and WebAssembly bindings for in-browser use. As an embedded database manager, it is easily integrated into applications. At Predictable Labs, Kuzu was a key component of a larger data system they are developing. For months, they have been working on their own fork, refining an earlier version of the codebase to meet specific requirements of their demanding use cases and workloads. The archiving of the original repository was a clear signal. To consistently deliver a robust, scalable, and reliable product to the enterprise market, they decided to mitigate risk by controlling the core graph component. Predictable Labs is now formally introducing RyuGraph, their open-source fork of KuzuDB, under the MIT licence. The decision to open source RyuGraph is driven by confidence that enterprise use cases will serve as the perfect proving ground, ensuring the project remains hardened and innovative. Predictable Labs brings deep, practical experience to RyuGraph, helping engineer it to meet a wide variety of production demands. Predictable Labs has been involved with the LEX and Text2Graph working group at the Graph Data Council (GDC) (formerly the Linked Data Benchmark Council) for the last couple of years, which they say reflects the belief that working collaboratively to establish standards and benchmarks is the most valuable way to contribute to the technology community. This moment has sparked activity across the community, leading to the creation of other forks. Predictable Labs claim that they will actively look for opportunities to collaborate for the common good. github.com/predictable-labs/… #OpenSource #Announcement -- The Year of the Graph's Autumn 2025 newsletter issue on all things #KnowledgeGraph, #GraphDB, Graph #Analytics / #DataScience / #AI and #SemTech is out. Subscribe and follow to be in the know. Reach out if you'd like to be featured 👇 yearofthegraph.xyz/newslette…

Make your medical AI smarter with better knowledge graphs. Lettria’s text2Graph shows you how.👇 lettria.com/lettria-lab/cons… #AI #KnowledgeGraphs #MedicalInnovation #Medicine

Make your medical AI smarter with better knowledge graphs. Lettria’s text2Graph shows you how.👇 lettria.com/lettria-lab/cons… #AI #KnowledgeGraphs #MedicalInnovation #Medicine

Want a smarter medical AI? Build better knowledge graphs. Here’s how Lettria’s text2Graph makes it happen.👇 lettria.com/lettria-lab/cons… #AI #KnowledgeGraphs #MedicalInnovation #Medicine

这个开源的 text2graph 的工具也太好了⚡️ 可以将文本或 url 信息结合 LLM 的常识自动转成可视化的知识图谱🔥 目前已经 3.3K Star 🌟 非常适合想要快速抓住一篇长文或者一个概念重点以及概念与概念之间关系的场景。 github.com/yoheinakajima/ins… 在线体验 👉 instagraph.ai/

LGGM: Large Graph Generative Models Text2Graph Generation Demo LGGM-Text2Graph is a large graph generative model that supports generating graphs from textual descriptions. #EmergingTech #OpenSource #DataViz #LLM #AI #DataScience #Analytics #GenAI buff.ly/4cN0uCB

Text2graph - free online tool from #Graphlytic for transformation of free text to linked graphs. With #LLM from #OpenAI and some clever features. See more in the video. youtube.com/watch?v=eChFhP_3…

Say hello to #text2graph that puts the power of data analysis in your hands, minus the coding. Effortlessly generate code for plots in R, Python and JS, customize models, and explore the stories hidden within your data 🚀 Want to learn more? Book a demo: go.appsilon.com/text2graph-a…

Check out our revamped #Text2graph tool. It uses #LLM from #OpenAI to detect entities and their relationships in free text. It's great for investigation or just playing around with film synopses :) Simply enter any text or drag&drop a file to start. graphlytic.com/blog/introduc…

Excellent Meetup on how the Semantic Web is meeting Industry today at BOPE Paris, meetup.com/graphes-semantiqu… Charles from @lettria_fr unveils the Text2Graph product or how to generate KG out of chaos of unstructured data! And of course, we talk about #LLM and #RAG based on graphs!

"InstaGraph" by @yoheinakajima let's you convert natural language prompts into a visualized knowledge graph (text2graph): instagraph.ai/

Bitcoin, visualized by #InstaGraph, text2graph AI agent. #cytoscape instagraph.ai/graph/oWFmeccV…

Here is the knowledge graph generated from the wikipedia article about you, using InstaGraph (text2graph AI agent). instagraph.ai/graph/oWFmeccV…

Knowledge graph about @ylecun, generated from the Wikipedia article by #InstaGraph, text2graph AI engine. instagraph.ai/graph/oWFmeccV…

Knowledge graph about @geoffreyhinton, generated from the wikipedia article using #InstaGraph, text2graph AI engine. instagraph.ai/graph/oWFmeccV…

Just deployed InstaGraph v.0.1.3, a "Text2Graph" AI agent, which generates a knowledge graph from text. @yoheinakajima #instagraph - It is now accessible to anybody (free) - Generate, Edit, Merge, Undo/Redo, Share, Remix ... instagraph.ai/

Here is the knowledge graph of OpenAI saga (simplified), generated by InstaGraph (GPT-based text2graph tool). instagraph.ai/graph/oWFmeccV…

Introducing TEXT2GRAPH - AI-powered online tool for quick text-to-graph transformation with a unique continuous learning mechanism. Create graphs from text paragraphs in seconds. #NLP #graphanalytics #cypher #gremlin #Datavisualization graphlytic.com/blog/introduc…