GraphMatcher System Presentation

Abstract

Ontology matching finds a relationship or correspondence between two or more entities in two or more ontologies. To solve the interoperability of the domain ontologies, semantically correspondence entities in these ontologies must be identified and aligned before merging them. GraphMatcher is an ontology matching system using a graph attention approach to compute higher-level representation of a class together with its surrounding terms.

Keywords

• eol>graph attention
• graph representation
• ontology matching

-

networks [ 5, 6, 4 ], to identify semantically corresponding concepts within the ontologies. The graph attention mechanism computes the higher-level representation of a concept and its surrounding concepts, and the model subsequently determines similarity scores between pairs of concepts.

1.2. Specific techniques used

GraphMatcher [ 1 ] utilizes a graph representation learning approach employing graph attention [ 2 ] and a supervised machine learning algorithm with a network consisting of five layers. The primary contribution is the adaptation of the graph attention mechanism to compute the higher-level representation of the contextual embedding of the center class. It is important to note that this year, we submitted the model that yielded the best performance results in its hyperparameter tuning process. There have been no other improvements or changes to the model.

1.3. Adaptations made for the evaluation

The GraphMatcher framework has been developed in Python using PyTorch and Ontospy, and it is packaged by SEALS using MELT [ 7 ]. However, because of the Universal Sentence Encoder’s TensorFlow Hub dependency, the model might not run on machines that do not support this library. In the case of this dependency issue, please use another sentence encoder.

1.4. Parameter settings

The model’s parameters include a learning rate of 0.01, five epochs, weight decay of 0.01, and a batch size of 32. The threshold is computed from false positive alignments in the validation data, following the approach proposed by the VeeAlign system [ 4 ] 1. These parameters were determined through five-fold cross-validation. The code is available at https://github.com/ sefeoglu/gat_ontology_matching.

2. Results

The results of the GraphMatcher in the OAEI 2023 conference tracks are available at https: //oaei.ontologymatching.org/2023/results/conference/.

3. General Comments

We plan to change the algorithm to an unsupervised machine learning approach in its next version, as there is no explicit rule regarding the supervised machine learning approach in the OAEI. In light of the study introducing the graph attention approach [ 2 ], this algorithm has also been adapted to the unsupervised approach. 1The project uses VeeAlign’s approach directly to compute the threshold with the permission of the first author.

References

[1] S. Efeoglu , Graphmatcher: A graph representation learning approach for ontology matching ( 2022 ).

[2] P. Veličković , G. Cucurull , A. Casanova , A. Romero , P. Liò , Y. Bengio , Graph attention networks ( 2018 ). arXiv: 1710 . 10903 .

[3] P. Kolyvakis , A. Kalousis , D. Kiritsis, DeepAlignment: Unsupervised ontology matching with refined word vectors , in: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , Volume 1 ( Long Papers) , ACL , 2018 , pp. 787 - 798 .

[4] V. Iyer1 , A. Agarwal , H. Kumar , Veealign: A supervised deep learning approach to ontology alignment ( 2020 ).

[5] A. Bento , A. Zouaq , M. Gagnon , Ontology matching using convolutional neural networks , in: Proceedings of the 12th Language Resources and Evaluation Conference , ELRA, Marseille, France, 2020 , pp. 5648 - 5653 . URL: https://www.aclweb.org/anthology/2020.lrec- 1 . 693 .

[6] J. Chen , J. Gu , Adol: a novel framework for automatic domain ontology learning , Super Computing ( 2021 ).

[7] S. Hertling , J. Portisch , H. Paulheim , Melt - matching evaluation toolkit , in: M. Acosta , P. Cudré-Mauroux , M. Maleshkova , T. Pellegrini , H. Sack , Y. Sure-Vetter (Eds.), Semantic Systems. The Power of AI and Knowledge Graphs , Springer International Publishing, Cham, 2019 , pp. 231 - 245 .