Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GraphMatcher: A Graph Representation Learning Approach for Ontology Matching

Sefika Efeoglu

TL;DR

GraphMatcher tackles ontology matching by encoding each concept's context via graph attention over an arbitrary graph representation of ontologies. It employs a Siamese architecture with a heterogeneous graph attention layer, neighbourhood aggregation, and Universal Sentence Encoder embeddings to produce contextual representations and cosine-based alignment scores. The key contributions include adapting graph attention to a Siamese setting over multiple homogeneous subgraphs and detailing a complete preprocessing and evaluation pipeline. Empirically, it yields strong F1 on M1 and M3 variants and high confidence on uncertain-reference tracks (around 72%), though it lags on M2 property alignments, with future work focusing on enriching property context and external information to improve this gap.

Abstract

Ontology matching is defined as finding a relationship or correspondence between two or more entities in two or more ontologies. To solve the interoperability problem of the domain ontologies, semantically similar entities in these ontologies must be found and aligned before merging them. GraphMatcher, developed in this study, is an ontology matching system using a graph attention approach to compute higher-level representation of a class together with its surrounding terms. The GraphMatcher has obtained remarkable results in in the Ontology Alignment Evaluation Initiative (OAEI) 2022 conference track. Its codes are available at ~\url{https://github.com/sefeoglu/gat_ontology_matching}.

GraphMatcher: A Graph Representation Learning Approach for Ontology Matching

TL;DR

GraphMatcher tackles ontology matching by encoding each concept's context via graph attention over an arbitrary graph representation of ontologies. It employs a Siamese architecture with a heterogeneous graph attention layer, neighbourhood aggregation, and Universal Sentence Encoder embeddings to produce contextual representations and cosine-based alignment scores. The key contributions include adapting graph attention to a Siamese setting over multiple homogeneous subgraphs and detailing a complete preprocessing and evaluation pipeline. Empirically, it yields strong F1 on M1 and M3 variants and high confidence on uncertain-reference tracks (around 72%), though it lags on M2 property alignments, with future work focusing on enriching property context and external information to improve this gap.

Abstract

Ontology matching is defined as finding a relationship or correspondence between two or more entities in two or more ontologies. To solve the interoperability problem of the domain ontologies, semantically similar entities in these ontologies must be found and aligned before merging them. GraphMatcher, developed in this study, is an ontology matching system using a graph attention approach to compute higher-level representation of a class together with its surrounding terms. The GraphMatcher has obtained remarkable results in in the Ontology Alignment Evaluation Initiative (OAEI) 2022 conference track. Its codes are available at ~\url{https://github.com/sefeoglu/gat_ontology_matching}.
Paper Structure (15 sections, 5 equations, 2 figures, 2 tables)

This paper contains 15 sections, 5 equations, 2 figures, 2 tables.

Table of Contents

  1. Presentation of the system
  2. Proposal and general statement
  3. Specific techniques used
  4. Preprocessing
  5. Embedding layer
  6. Heterogeneous graph attention layer
  7. Output layer
  8. Cosine similarity layer
  9. Adaptations made for the evaluation
  10. Parameter settings
  11. Results
  12. General Comments
  13. Comments on the results
  14. Improvements
  15. Conclusion

Figures (2)

  • Figure 1: The proposed network is an application of heterogeneous graph attention on Siamese networks to find the similar classes.
  • Figure 2: This figure shows how graph attention is applied to a homogeneous graph. The relationship between the centre class and its neighbouring classes is the same in the homogeneous graphs such as 'subClassOf'. In this graph, the 'ObjectProperty' node represents the restrictions belonging to the definition of a child or parent class. Bag-Of-Word is used to represent the centre class and its neighbouring classes. $h_{1}^{'}$ shows the contextual information of $h_{1}$ in terms of this 'subClassOf' relation after applying graph attention. In addition, a$_{ij}$ denotes $\alpha_{ij}$ in Equation \ref{['eq:gat_5']}.