Arabic ontology learning using deep learning

Abstract

Ontology, the backbone of Semantic Web, is defined as the formal specification of conceptual hierarchy with relationships between concepts. Ontology Learning (OL) is a process to create an ontology from text automatically or semi-automatically. OL is an important topic in the Semantic Web field in the last two decades but it is still not mature in Arabic not like Latin languages. Currently, there is a limited support for using knowledge from Arabic literature automatically in semantically-enabled systems. Deep Learning (DL), an artificial neural networks learning based application, has proved a good improvement in multiple areas including text mining. By using DL, it is possible to have word embedding as distributed word representations from textual data. The application of DL to aid Arabic ontology development remains largely unexplored. This paper investigates the performance of implementing DL with Arabic ontology learning tasks using major models such as Continuous Bag of Words (CBOW) and Skip-gram. Initial performance results are promising as an effective application of Arabic ontology learning.