An Ontology of Tobacco Production: Enriching Large Language Model-based Decision Support

Tobacco (Nicotiana tabacum) production plays a crucial role in the agricultural economy of several regions around the world, especially in developing countries, such as Brazil. However, the lack of well-defined terminology and semantic models harms the development of decision support systems. Conceptualizing the tobacco production domain is challenging due to its ambiguous terminology and the complexity involved in considering environmental, soil, disease, and pest management factors. We present an Ontology of Tobacco Production (OnTop), designed to assist tobacco production in optimizing crop management practices and placing environmental factors within a formal framework. To the best of available knowledge, this study is the first to formalize the tobacco production lifecycle integrated with soil and climate data. It includes symbolic and description logic for reasoning and automation. This paper presents the proposal for the core set (main classes) and the application of OnTop to enrich a Large Language Model (LLM) to provide an ontology-based decision support prompt. OnTop allows actionable recommendations based on environmental, agronomic, and productivity data. The main contributions of this paper are: i) a domain ontology that formalizes knowledge on data-driven tobacco production; and ii) a queryable framework (enriched LLM), which allows experts to obtain complex agronomic answers. OnTop offers an extensible framework for decision making in tobacco farming and paves the way for further innovations in ontologies and LLM-based decision support systems for the agricultural domain.