Mining Knowledge from Data: The Case of Ontology Learning

For many years, ontologies have set a standard for describing and representing various domains, capturing concepts and their relationship, and improving knowledge management. Nowadays, ontologies are often created manually, although various attempts have been made in the literature to automatically generate them (Ontology Learning). An ontology could be generated starting from different sources of data and types of databases: although SQL databases remain the most popular in the market today, NoSQL databases are carving out a large share of the market as they possess numerous advantages over their competitors (for example, flexible data models, faster queries, etc.). This article reviews the main Ontology Learning methods available in the literature that can be applied to NoSQL databases. Such methods will be categorized based on the central NoSQL database (Document-oriented, Column-family-oriented, Key-value, and Graph databases), which can be the data source of an ontology learning method. Also, methods for Ontology Learning starting from JSON documents are considered. The advantages and disadvantages of the approaches have been highlighted and summarized to assess the state of the art in the field of NoSQL ontology learning and possible improvements and future work in such context.