Predicting the top-level ontological concepts of domain entities using wordembeddings, informal definitions, and deep learning

Ontology development is a challenging task that encompasses many time-consuming activities. One of theseactivities is the classification of the domain entities (concepts and instances) according to top-level concepts.This activity is usually performed manually by an ontology engineer. However, when the set of entitiesincreases in size, associating each entity to the proper top-level ontological concept becomes challenging andrequires a high level of expertise in both the target domain and ontology engineering. This paper proposes adeep learning approach that automatically classifies domain entities into top-level concepts using their informaldefinitions and the word embedding of the terms that represent them. From these inputs, we feed a deepneural network consisting of two modules: a feed-forward neural network and a bi-directional recurrent neuralnetwork with long short-term units. Our architecture combines both outputs of these modules into a denselayer and provides the probabilities of each candidate class. For validating our proposal, we have developed adataset based on the OntoWordNet ontology, which provides a classification of WordNet synsets into conceptsspecified by DOLCE-lite-plus top-level ontology. Our experiments show that our proposal outperforms thebaseline approaches by 6% regarding the F-score. In addition, our proposal is less affected by the polysemy inthe terms that represent the domain entities than the compared approaches. Consequently, our proposal canconsider more instances during its training than the baseline methods