GraphMatcher

The GraphMatcher aims to find the correspondes between two ontologies and outputs the possible alignments between them.

The GraphMatcher leverages Graph Attention Network[2] in its neural network structure. The project leverages a new neighborhood aggregation algorithm, so it examines contribution of neighboring terms which have not been used in the previous matchers before.

The project has been submitted to The 17th International Workshop on Ontology Matching's OAEI 2022 (ISWC-2022) for conference track and obtained the highest F1-measure in uncertain reference alignments among other experts participating to this challenge. Its system paper has been published, and it was invited to the poster presentation session.

Set up

• 1.) install requirements pip install -r requirements.txt

• 2.) set the parameters in the config.ini

[General]
dataset =               ------> name of a dataset e.g., conference.
K =                     ------> the parameter for K fold cross-validation
ontology_split =        ------> True/False
max_false_examples =    ------>

[Paths]
dataset_folder =        ------> a path to the ontologies
alignment_folder =      ------> a path to the reference alignments
save_model_path =       ------> save the model to the path
load_model_path =       ------> model path
output_folder =         ------> The output folder for the alignments

[Parameters]
max_paths =             ------>
max_pathlen =           ------> ( number of neighboring concepts' types: Equivalent class, subclass of(general to specific or specific to general(2))...
[Hyperparameters] 

lr =                    ------> learning rate
num_epochs =            ------> number of epochs
weight_decay =          ------> Weight decay
batch_size =            ------> Batch Size (8/16/32)

• 3.) train the model

python src/train_model.py

• 4.) test the model

python src/test_model.py ${source.rdf} ${target.rdf}

Sample Alignment:

  <map>
    <Cell>
      <entity1 rdf:resource='http://conference#has_the_last_name'/>
      <entity2 rdf:resource='http://confof#hasSurname'/>
      <relation>=</relation>
      <measure rdf:datatype='http://www.w3.org/2001/XMLSchema#float'>0.972</measure>
    </Cell>
  </map>

• 5.) evaluate the model with the MELT

Note: The codes in train_model.py and test_model.py are partially based on the VeeAlign[2] project with the permission of its main author. I would like to thank the main author.

References:

[1]

@inproceedings{iyer-etal-2021-veealign,
    title = "{V}ee{A}lign: Multifaceted Context Representation Using Dual Attention for Ontology Alignment",
    author = "Iyer, Vivek  and
      Agarwal, Arvind  and
      Kumar, Harshit",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-main.842",
    doi = "10.18653/v1/2021.emnlp-main.842",
    pages = "10780--10792",
   }

[2]

@misc{https://doi.org/10.48550/arxiv.1710.10903,
  title = {Graph Attention Networks},
  author = {Veličković, Petar and Cucurull, Guillem and Casanova, Arantxa and Romero, Adriana and Liò, Pietro and Bengio, Yoshua},
  keywords = {Machine Learning (stat.ML), Artificial Intelligence (cs.AI), Machine Learning (cs.LG), Social and Information Networks (cs.SI), FOS: Computer and information sciences, FOS: Computer and information sciences},
  url = {https://arxiv.org/abs/1710.10903},
  publisher = {arXiv},
  doi = {10.48550/ARXIV.1710.10903},
  year = {2017},
  copyright = {arXiv.org perpetual, non-exclusive license}
}