With the progressive increase in the number of existing ontologies, ontology matching became a challenging task. Ontology matching is a crucial step in the ontology integration process and its goal is to find correspondent elements in heterogeneous ontologies. A trend of clustering-based solutions for ontology matching has evolved, based on a divide-and-conquer strategy, which partitions ontologies, clusters similar partitions and restricts the matching to ontology elements of similar partitions. Nevertheless, most of these solutions considered solely the terminological aspect, ignoring other ontology aspects that can contribute to the final matching results. In this work, we developed a novel solution for ontology matching based on a consensus clustering of multiple aspects of ontology partitons. We partitioned the ontologies applying Community Detection techniques and applied Bayesian Cluster Ensembles (BCE) to find a consensus clustering among the terminological, topological and extensional aspects of ontology partitions. The matching results of our experimental study indicated that a BCE-based solution with three clusters best captured the contributions of the aspects, in comparison to other consensual solutions. The results corroborated the benefits of the synergy between the ontology aspects to the ontology alignment. We also verified that the BCE-based solution for three clusters yielded higher matching scores than other state-of-the-art solutions. Besides, our proposed methods structurize a configurable framework, which allows adding other ontology aspects and also other techniques.

中文翻译：

---

一种基于贝叶斯聚类集成的本体匹配多方面方法

随着现有本体数量的逐渐增加，本体匹配成为一项具有挑战性的任务。本体匹配是本体集成过程中的关键步骤，其目标是在异构本体中找到对应的元素。一种基于聚类的本体匹配解决方案的发展趋势是基于分而治之的策略，即划分本体，聚类相似的分区，并将匹配限制在相似分区的本体元素上。尽管如此，这些解决方案中的大多数只考虑了术语方面，而忽略了可能有助于最终匹配结果的其他本体方面。在这项工作中，我们基于本体分区的多个方面的共识聚类开发了一种新的本体匹配解决方案。我们应用社区检测技术对本体进行划分，并应用贝叶斯聚类集成（BCE）来找到本体划分的术语、拓扑和扩展方面之间的共识聚类。我们的实验研究的匹配结果表明，与其他共识解决方案相比，基于 BCE 的三个集群的解决方案最能捕捉到这些方面的贡献。结果证实了本体方面之间的协同作用对本体对齐的好处。我们还验证了基于 BCE 的三个集群的解决方案比其他最先进的解决方案产生了更高的匹配分数。此外，我们提出的方法构建了一个可配置的框架，允许添加其他本体方面和其他技术。