Automating Computer Science Ontology Extension With Classification Techniques

Abstract

In information technology, an ontology is a knowledge structure consisting of the definitions and relations of information within one or even multiple domains. This semantically represented information is helpful for tasks such as document classification and item recommendation in recommender systems. However, as big data prevails, manually extending existing ontologies with up-to-date terminologies becomes challenging due to the tedious and time-consuming process and the expensive cost of expert manual labor. This study aims to achieve a fully automatic ontology extension.We propose a novel "Direct" approach for extending an existing Computer Science Ontology (CSO). This approach consists of two steps: initially extending the CSO with new topics and using this extended graph to obtain the new topic’s node embeddings as inputs for training classifiers. However, this initial extension still contains many noisy links; therefore, the classifier later acts as a filter and a link predictor. We experiment with various traditional machine learning and recent deep learning models and then compare them using our Direct approach. We also propose two evaluation procedures to decide the best-performing model and approach: the novel Wikipedia-based F1w score and the total number of resulting links. Furthermore, manual evaluation by four human experts is conducted to conclude the reliability of our proposed approach and evaluation procedure. This study concludes that the Direct approach’s Gaussian Naive Bayes model produces the most valid and reliable links, and we, therefore, use it to further extend the CSO with hundreds of new CS topics and links.