Deep Learning-Driven Ontology Learning: A Systematic Mapping Study

Abstract

Today, ontologies are the widely accepted framework for managing knowledge in a manner that supports sharing, reuse, and automatic interpretation. Ontologies are fundamental to various Artificial Intelligence (AI) applications, including smart information retrieval, knowledge management, and contextual organization. However, the rapid growth of data in various domains has made ontology acquisition and enrichment, time-consuming, labor-intensive, and expensive. Consequently, there is a need for automated methods for this task, commonly referred to as ontology learning. Deep learning models have made significant advancements in this field, as they can extract concepts from vast corpora and infer semantic relationships from wide-ranging datasets. This paper aims to explore and synthesize existing research on the application of deep learning techniques to ontology learning. To achieve this, a Systematic Mapping Study (SMS) was conducted, encompassing 2765 papers published between 2015 and September 2024, from which 47 research papers were selected for review and analysis. The studies were systematically categorized according to eight refined criteria: publication year, type of contribution, empirical study design, type of data used, deep learning techniques implemented, domain of application, focused ontology learning tasks, and evaluation metrics and benchmarks.