Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CE-GOCD: Central Entity-Guided Graph Optimization for Community Detection to Augment LLM Scientific Question Answering

Jiayin Lan, Jiaqi Li, Baoxin Wang, Ming Liu, Dayong Wu, Shijin Wang, Bing Qin, Guoping Hu

TL;DR

CE-GOCD addresses the challenge of robust scientific question answering with LLMs by explicitly modeling semantic substructures in a knowledge graph of academic papers. It introduces a title-centered pipeline: (i) subgraph retrieval using center entities and lexical cues, (ii) subgraph pruning and completion to uncover implicit relations, and (iii) title-guided community detection via modularity optimization to group thematically related papers. The method yields superior QA performance on NLP benchmarks and demonstrates domain generalizability to medical data, with ablations confirming the necessity of both subgraph optimization and community-level reasoning. The approach offers a principled way to augment LLM QA with structured, thematically coherent evidence across papers, enhancing accuracy and interpretability in scientific literature search and reasoning.

Abstract

Large Language Models (LLMs) are increasingly used for question answering over scientific research papers. Existing retrieval augmentation methods often rely on isolated text chunks or concepts, but overlook deeper semantic connections between papers. This impairs the LLM's comprehension of scientific literature, hindering the comprehensiveness and specificity of its responses. To address this, we propose Central Entity-Guided Graph Optimization for Community Detection (CE-GOCD), a method that augments LLMs' scientific question answering by explicitly modeling and leveraging semantic substructures within academic knowledge graphs. Our approach operates by: (1) leveraging paper titles as central entities for targeted subgraph retrieval, (2) enhancing implicit semantic discovery via subgraph pruning and completion, and (3) applying community detection to distill coherent paper groups with shared themes. We evaluated the proposed method on three NLP literature-based question-answering datasets, and the results demonstrate its superiority over other retrieval-augmented baseline approaches, confirming the effectiveness of our framework.

CE-GOCD: Central Entity-Guided Graph Optimization for Community Detection to Augment LLM Scientific Question Answering

TL;DR

CE-GOCD addresses the challenge of robust scientific question answering with LLMs by explicitly modeling semantic substructures in a knowledge graph of academic papers. It introduces a title-centered pipeline: (i) subgraph retrieval using center entities and lexical cues, (ii) subgraph pruning and completion to uncover implicit relations, and (iii) title-guided community detection via modularity optimization to group thematically related papers. The method yields superior QA performance on NLP benchmarks and demonstrates domain generalizability to medical data, with ablations confirming the necessity of both subgraph optimization and community-level reasoning. The approach offers a principled way to augment LLM QA with structured, thematically coherent evidence across papers, enhancing accuracy and interpretability in scientific literature search and reasoning.

Abstract

Large Language Models (LLMs) are increasingly used for question answering over scientific research papers. Existing retrieval augmentation methods often rely on isolated text chunks or concepts, but overlook deeper semantic connections between papers. This impairs the LLM's comprehension of scientific literature, hindering the comprehensiveness and specificity of its responses. To address this, we propose Central Entity-Guided Graph Optimization for Community Detection (CE-GOCD), a method that augments LLMs' scientific question answering by explicitly modeling and leveraging semantic substructures within academic knowledge graphs. Our approach operates by: (1) leveraging paper titles as central entities for targeted subgraph retrieval, (2) enhancing implicit semantic discovery via subgraph pruning and completion, and (3) applying community detection to distill coherent paper groups with shared themes. We evaluated the proposed method on three NLP literature-based question-answering datasets, and the results demonstrate its superiority over other retrieval-augmented baseline approaches, confirming the effectiveness of our framework.
Paper Structure (14 sections, 6 equations, 3 figures, 4 tables)

This paper contains 14 sections, 6 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. PROPOSED METHOD CE-GOCD
  3. Subgraph Retrieval
  4. Subgraph Optimization
  5. Community Detection
  6. EXPERIMENT
  7. Relevant Dataset
  8. Implementation Details
  9. Experimental Results
  10. Domain Generalizability Experimental Results
  11. Ablation Experimental Results
  12. Graph Optimization Bias Analysis
  13. Conclusions
  14. Acknowledgment

Figures (3)

  • Figure 1: Knowledge graph augmentation for scientific QA.
  • Figure 2: (a) CE-GOCD chematic diagram. We perform graph retrieval (b) using LLM-based filtering along with inter-entity path and neighbor subgraph extraction. This is followed by subgraph optimization and community detection (c), which involve graph pruning, graph completion, and community-level answers generation.
  • Figure 3: An example of pruning (a) and completion (b).