Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Panning for Gold: Expanding Domain-Specific Knowledge Graphs with General Knowledge

Runhao Zhao, Weixin Zeng, Wentao Zhang, Chong Chen, Zhengpin Li, Xiang Zhao, Lei Chen

TL;DR

We address the problem of enriching domain-specific knowledge graphs with facts from general knowledge graphs. The authors propose ExeFuse, a Fact-as-Program framework where each general KG fact is treated as a latent semantic program; by mapping abstract relations to granularity-aware operators and verifying executability on the target DKG, the method unifies relevance detection and granularity transformation. They formalize a probabilistic fusion objective and train a multi-stage model that first constructs executable programs, then runs them on the DKG, and finally verifies executability before predicting fused facts. To enable systematic evaluation, they introduce two benchmarks, DKGF(W-I) and DKGF(Y-I), with 21 configurations, and provide extensive experiments showing improvements over baselines, transferability to unseen entities, efficiency advantages, and insight into the role of relevant entity finding. The work offers a practical, scalable pathway for leveraging rich GKGs to enrich specialized domains, with potential impact on domain-specific reasoning, crisis analysis, and knowledge-driven applications.

Abstract

Domain-specific knowledge graphs (DKGs) often lack coverage compared to general knowledge graphs (GKGs). To address this, we introduce Domain-specific Knowledge Graph Fusion (DKGF), a novel task that enriches DKGs by integrating relevant facts from GKGs. DKGF faces two key challenges: high ambiguity in domain relevance and misalignment in knowledge granularity across graphs. We propose ExeFuse, a simple yet effective Fact-as-Program paradigm. It treats each GKG fact as a latent semantic program, maps abstract relations to granularity-aware operators, and verifies domain relevance via program executability on the target DKG. This unified probabilistic framework jointly resolves relevance and granularity issues. We construct two benchmarks, DKGF(W-I) and DKGF(Y-I), with 21 evaluation configurations. Extensive experiments validate the task's importance and our model's effectiveness, providing the first standardized testbed for DKGF.

Panning for Gold: Expanding Domain-Specific Knowledge Graphs with General Knowledge

TL;DR

We address the problem of enriching domain-specific knowledge graphs with facts from general knowledge graphs. The authors propose ExeFuse, a Fact-as-Program framework where each general KG fact is treated as a latent semantic program; by mapping abstract relations to granularity-aware operators and verifying executability on the target DKG, the method unifies relevance detection and granularity transformation. They formalize a probabilistic fusion objective and train a multi-stage model that first constructs executable programs, then runs them on the DKG, and finally verifies executability before predicting fused facts. To enable systematic evaluation, they introduce two benchmarks, DKGF(W-I) and DKGF(Y-I), with 21 configurations, and provide extensive experiments showing improvements over baselines, transferability to unseen entities, efficiency advantages, and insight into the role of relevant entity finding. The work offers a practical, scalable pathway for leveraging rich GKGs to enrich specialized domains, with potential impact on domain-specific reasoning, crisis analysis, and knowledge-driven applications.

Abstract

Domain-specific knowledge graphs (DKGs) often lack coverage compared to general knowledge graphs (GKGs). To address this, we introduce Domain-specific Knowledge Graph Fusion (DKGF), a novel task that enriches DKGs by integrating relevant facts from GKGs. DKGF faces two key challenges: high ambiguity in domain relevance and misalignment in knowledge granularity across graphs. We propose ExeFuse, a simple yet effective Fact-as-Program paradigm. It treats each GKG fact as a latent semantic program, maps abstract relations to granularity-aware operators, and verifies domain relevance via program executability on the target DKG. This unified probabilistic framework jointly resolves relevance and granularity issues. We construct two benchmarks, DKGF(W-I) and DKGF(Y-I), with 21 evaluation configurations. Extensive experiments validate the task's importance and our model's effectiveness, providing the first standardized testbed for DKGF.
Paper Structure (33 sections, 11 equations, 4 figures, 5 tables)

This paper contains 33 sections, 11 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Integration from Structured Knowledge
  4. Entity alignment
  5. Knowledge Graph Completion
  6. Integration from Unstructured Knowledge
  7. Relation Triple Extraction
  8. Entity Linking
  9. Problem statement
  10. Domain-specific Knowledge Graph
  11. Domain-specific Knowledge Graph Fusion
  12. Methodology
  13. Core Insight: Relevance as Executability
  14. Unified Probabilistic Framework
  15. Model Architecture
  16. ...and 18 more sections

Figures (4)

  • Figure 1: Examples of domain-specific knowledge graph fusion. \ref{['fig:intro']}(b) demonstrates the fusion of highly relevant entities from GKG into DKG. \ref{['fig:intro']}(c) shows the augmented DKG.
  • Figure 2: The framework of ExeFuse for DKGF.
  • Figure 3: The results of the transferability experiment on the DKGF(W-I) and DKGF(Y-I) datasets at different training rates. The test data is split into seen (S) and unseen (U) parts based on whether the entity appeared during training. The total results for all (A) test data are also reported. Accuracy (ACC) and F1-score (F1) are shown in the radar charts.
  • Figure 4: The venn diagram of correct predictions across various DKGF benchmark configurations on DKGF(W-I)-S1. Intersections show shared same predictions.