Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Customized Information and Domain-centric Knowledge Graph Construction with Large Language Models

Frank Wawrzik, Matthias Plaue, Savan Vekariya, Christoph Grimm

TL;DR

The paper addresses the need for timely, machine-actionable access to technology intelligence to plan cyber-physical systems. It introduces a two-stage knowledge graph construction framework that pre-selects domain-relevant documents from heterogeneous sources and anchors the KG to the GENIAL! Basic Ontology, enabling domain-specific reasoning. The approach is demonstrated in automotive electrical systems, showing improvements over baselines in class recognition and relationship extraction, and offering a scalable alternative to Wikidata for domain content. The results yield rich topic maps and knowledge graphs that support planning and decision-making, with future work focusing on quality assurance and ontology-driven refinement.

Abstract

In this paper we propose a novel approach based on knowledge graphs to provide timely access to structured information, to enable actionable technology intelligence, and improve cyber-physical systems planning. Our framework encompasses a text mining process, which includes information retrieval, keyphrase extraction, semantic network creation, and topic map visualization. Following this data exploration process, we employ a selective knowledge graph construction (KGC) approach supported by an electronics and innovation ontology-backed pipeline for multi-objective decision-making with a focus on cyber-physical systems. We apply our methodology to the domain of automotive electrical systems to demonstrate the approach, which is scalable. Our results demonstrate that our construction process outperforms GraphGPT as well as our bi-LSTM and transformer REBEL with a pre-defined dataset by several times in terms of class recognition, relationship construction and correct "sublass of" categorization. Additionally, we outline reasoning applications and provide a comparison with Wikidata to show the differences and advantages of the approach.

Customized Information and Domain-centric Knowledge Graph Construction with Large Language Models

TL;DR

The paper addresses the need for timely, machine-actionable access to technology intelligence to plan cyber-physical systems. It introduces a two-stage knowledge graph construction framework that pre-selects domain-relevant documents from heterogeneous sources and anchors the KG to the GENIAL! Basic Ontology, enabling domain-specific reasoning. The approach is demonstrated in automotive electrical systems, showing improvements over baselines in class recognition and relationship extraction, and offering a scalable alternative to Wikidata for domain content. The results yield rich topic maps and knowledge graphs that support planning and decision-making, with future work focusing on quality assurance and ontology-driven refinement.

Abstract

In this paper we propose a novel approach based on knowledge graphs to provide timely access to structured information, to enable actionable technology intelligence, and improve cyber-physical systems planning. Our framework encompasses a text mining process, which includes information retrieval, keyphrase extraction, semantic network creation, and topic map visualization. Following this data exploration process, we employ a selective knowledge graph construction (KGC) approach supported by an electronics and innovation ontology-backed pipeline for multi-objective decision-making with a focus on cyber-physical systems. We apply our methodology to the domain of automotive electrical systems to demonstrate the approach, which is scalable. Our results demonstrate that our construction process outperforms GraphGPT as well as our bi-LSTM and transformer REBEL with a pre-defined dataset by several times in terms of class recognition, relationship construction and correct "sublass of" categorization. Additionally, we outline reasoning applications and provide a comparison with Wikidata to show the differences and advantages of the approach.
Paper Structure (23 sections, 10 figures, 3 tables)

This paper contains 23 sections, 10 figures, 3 tables.

Table of Contents

  1. Introduction
  2. State of the Art
  3. Text mining for technology intelligence
  4. Topic maps and semantic networks
  5. Knowledge graph construction
  6. Planning and electronics ontologies
  7. Method
  8. Framework for technology intelligence andpPlanning - overview and approach
  9. Innovation Graph database
  10. Information retrieval and keyphrase extraction
  11. Semantic network
  12. Ontology for Innovation and Planning
  13. Evaluation
  14. Use Case: Automotive Electrical Systems Motivation
  15. LLMs to create semantic networks for data analytics
  16. ...and 8 more sections

Figures (10)

  • Figure 1: Pre-refined knowledge graph construction framework for innovation planning
  • Figure 2: Ontology for innovation planning
  • Figure 3: Example of a cluster within our semantic network "automotive electrical systems"
  • Figure 4: Search strategy for matching documents relevant to "automotive electrical systems"
  • Figure 5: Example cluster within a semantic network generated by prompting ChatGPT
  • ...and 5 more figures