knowledge-graph

This paper presents WaterBERT, a domain-adapted BERT model for water treatment literature mining, enhancing semantic representation and enabling large-scale structured information extraction and knowledge graph construction.

VisPath introduces a visual-intent-guided path reasoning framework for multimodal knowledge graph question answering, achieving significant improvements over baselines on a new benchmark, VisPath-Bench, and existing datasets.

HERMES is a graph-based framework that constructs personalized knowledge graphs from clinical notes using LLM-guided extraction and contrastive logic modeling, and applies graph attention networks for improved patient outcome prediction, outperforming text-only baselines on MIMIC datasets.

The UN System Data Commons is an open-source, AI-ready platform launched by Google and the UN that integrates global statistics into a single searchable resource, enabling natural language queries for easier data exploration.

PunGraph is a retrieval-enhanced knowledge graph framework that improves pun understanding by integrating phonetic and semantic constraints for large language models. It introduces a new dataset, WebPun, and demonstrates competitive performance on pun interpretation tasks.

OdoBot is a novel web-agent architecture that uses application behavior modeling to reduce token consumption and improve task success rates, outperforming agents like Agent-E and WebVoyager on the Canvas LMS.

DARE is a dialectical agentic reasoning framework for structured knowledge fact checking that uses an iterative retrieve–reason–reflect process, achieving 88.12% accuracy with an 8B model and matching GPT-4o performance.

This paper proposes a context-augmented training framework for multi-hop question-answering, showing that combining context graphs with knowledge graphs and using reinforcement learning improves performance in biomedical domains.

This paper presents data stories as narrative documents combining explanatory text, images, and executable SPARQL queries to enhance accessibility and quality assessment in cultural heritage knowledge graphs, using an AI-supported authoring platform called LODEON.

This paper introduces a framework for multilingual multimodal entity linking that improves accuracy on rare entities by combining reasoning and retrieval, with significant gains on the MERLIN benchmark.

This paper introduces ReTA, a reinforcement learning-based framework for dynamically augmenting electronic health record graphs with external knowledge graphs to improve prediction tasks like diagnosis and mortality.

This article presents SSAKG 2.0, an open-source software package for constructing and operating Structural Sequential Associative Knowledge Graphs (SSAKGs) for context-based sequence retrieval, with new efficient algorithms implemented in Python and C.

The author discusses inefficiencies in agents walking knowledge graphs for memory and proposes IWE, a tool that consolidates graph traversal into a single deterministic call to improve speed and reliability.

The paper proposes MAGG, a multi-agent framework for constructing governed knowledge graphs with domain-expert review, improving knowledge extraction and question-answering performance over methods like Microsoft GraphRAG.

This paper utilizes NLP techniques such as NER and BERTopic, along with Neo4j, to extract, classify, and visualize knowledge from translated ancient Indian medical texts, enhancing accessibility and digital preservation.

GitNexus is an open-source tool that creates a knowledge graph of codebases to help AI coding agents understand code relationships, significantly boosting their performance on benchmarks.

SNAP-KG is a framework that enables efficient integration of streaming entities into knowledge graphs via multi-view clustering and inductive inference, reducing inference time and candidate search space for downstream tasks like entity resolution and link prediction.

SelfGraphRAG introduces a framework that generates synthetic question-answer pairs from knowledge graphs to address the supervision gap in graph-based retrieval-augmented generation, enhancing retrieval precision and reasoning performance.

Understand Anything is an open-source tool that uses a multi-agent pipeline to analyze codebases and build interactive knowledge graphs, helping developers understand complex projects. It integrates with AI tools like Claude Code, Codex, Cursor, and has gained over 80,000 stars on GitHub.

An interactive knowledge graph builder shared by @yoheinakajima, featuring a playful metaphor involving meatballs and spaghetti for visualization or demonstration.