Graphusion: A RAG Framework for Scientific Knowledge Graph Construction with a Global Perspective

Graphusion: фреймворк RAG для построения научных графов знаний с глобальной перспективой
Jinghui Lu, Rui Yang, Irene Li, Tianwei She, Freddy Lécué, Xinjie Zhao, Boming Yang, Sixun Ouyang, Aosong Feng, Yuang Jiang, Moritz Blum, Fan Gao
2025-05-08

GraphusionTutorQA benchmarkglobal knowledge fusionknowledge graph constructionretrieval-augmented generation
Knowledge Graphs (KGs) are crucial in the field of artificial intelligence and are widely used in downstream tasks, such as questionanswering (QA).The construction of KGs typically requires significant effort from domain experts.Large Language Models (LLMs) have recently been used for Knowledge Graph Construction (KGC).However, most existing approaches focus on a local perspective, extracting knowledge triplets from individual sentences or documents, missing a fusion process to combine the knowledge in a global KG.This work introduces Graphusion, a zero-shot KGC framework from free text.It contains three steps: in Step 1, we extract a list of seed entities using topic modeling to guide the final KG includes the most relevant entities; in Step 2, we conduct candidate triplet extraction using LLMs; in Step 3, we design the novel fusion module that provides a global view of the extracted knowledge, incorporating entity merging, conflict resolution, and novel triplet discovery.Results show that Graphusion achieves scores of 2.92 and 2.37 out of 3 for entity extraction and relation recognition, respectively.Moreover, we showcase how Graphusion could be applied to the Natural Language Processing (NLP) domain and validate it in an educational scenario.Specifically, we introduce TutorQA, a new expert-verified benchmark for QA, comprising six tasks and a total of 1,200 QA pairs.Using the Graphusion-constructed KG, we achieve a significant improvement on the benchmark, for example, a 9.2% accuracy improvement on sub-graph completion.
1
A Graphusion-constructed knowledge graph improves benchmark performance, including a 9.2% accuracy gain on sub-graph completion.
2
Graphusion achieves scores of 2.92/3 for entity extraction and 2.37/3 for relation recognition.
3
Graphusion is a zero-shot framework for constructing knowledge graphs from free text using seed-entity extraction, LLM-based triplet extraction, and global knowledge fusion.
4
Its fusion module provides a global perspective by merging entities, resolving conflicts, and discovering novel triplets across extracted knowledge.
5
The authors introduce TutorQA, an expert-verified educational QA benchmark with six tasks and 1,200 question-answer pairs.

free-text scientific knowledge graphs constructed with Graphusion

global knowledge fusion for knowledge graph construction, including seed-entity selection, entity merging, conflict resolution, and novel triplet discovery

Publication Details
Publication Date
2025-05-08
Journal
Publisher
ISSN
Cited by
25
Access Type
Author Information
Authors
Jinghui Lu
Rui Yang
Irene Li
Tianwei She
Freddy Lécué
Xinjie Zhao
Boming Yang
Sixun Ouyang
Aosong Feng
Yuang Jiang
Moritz Blum
Fan Gao
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%