Zusammenfassung
Automating ontology curation is a crucial task in knowledge engineering.
Prediction by machine learning techniques such as semantic embedding is a
promising direction, but the relevant research is still preliminary. In this
paper, we present a class subsumption prediction method named BERTSubs, which
uses the pre-trained language model BERT to compute contextual embeddings of
the class labels and customized input templates to incorporate contexts of
surrounding classes. The evaluation on two large-scale real-world ontologies
has shown its better performance than the state-of-the-art.
Nutzer