From Manual to LLM-assisted: A Mapping Review on Ontology Construction Methods

Ontologies play a fundamental role in knowledge engineering and artificial intelligence tasks by providing a shared and formal representation of knowledge. Ontologies allow for a common understanding of concepts, support logical reasoning, and enable tasks such as inconsistency detection or instance checking. Formal ontologies pave the ground for knowledge reuse and sharing, and can be queried through dedicated query languages (SPARQL). Constructing ontologies is pivotal to describing a domain before instantiating it, or to generalizing from existing data. In the literature, plenty of ontology construction methods have been proposed, according to various scenarios and data. Overall, methods can be divided according to the nature of the process: manual, semi-automatic, or automatic. This distinction is nowadays becoming blurrier with LLMs playing a primary role in tempering their distance. In this plethora, choosing a suitable methodology according to the many factors involved is demanding. The goal of this work is therefore to put into context recent updates in ontology engineering and present the necessary additional information targeted to reuse. No review of manual ontology construction has been available for the last five years, and no classification of construction methods is available. This review fills this gap by providing an updated overview and proposing a two-tier categorization. The state-of-the-art is primarily divided into three categories (manual, semi-automatic, and automatic); for each category, representative features are then proposed for classification and comparison. To the best of our knowledge, this is the first review to offer such an intra-category classification.