A Framework for Enhanced Multi-Concept Based MaCOnto Ontology Evolution

Abstract

The discovery of ontology as semantic knowledge representation technique in solving the problems of unstructured data, concept mismatch is no doubt promising. However, the process of developing ontology manually based on the top down design approach is not just tedious and time consuming, but makes the ontology files static. Attempts have been made to solve this limitation by literature but not without some gaps. MaCOnto, a maize crop ontology attempted to solve the problem of ontology dynamism, but only one concept can autonomously evolved. Therefore, this research proposes a robust framework that can enable multiple related concepts of an existing ontology to evolve autonomously, a case study of MaCOnto in this research. The proposed design leverages on machine learning and deep learning techniques. The framework facilitates the autonomous evolution of the MaCOnto ontology through effective concept extraction and relationship mapping from diverse data sources, such as Wikipedia. Additionally, the research utilized the Term Frequency-Inverse Document Frequency (TF-IDF) for extracting domain-related concepts and Word2Vec algorithms for generating contextual word embeddings. The findings emphasize the significance of developing adaptive ontologies capable of evolving with dynamic knowledge domains, thus improving knowledge representation and reasoning in agricultural applications. Results indicate that the enhanced MaCOnto framework successfully achieved an average performance success rate of 83.5% across various competency questions, demonstrating its effectiveness in autonomously encoding multiple concepts and improving overall ontology functionality.