Biomedical Ontology Mapping with Green AI and Deep Learning

The integration of biomedical ontologies is a strategic element for public health and national industry, because information systems operate on preexisting, autonomous, distributed, and heterogeneous databases. In this context, the diversity of organizations in the biomedical domain must be considered when supporting business processes, which requires the management and development of these systems in a globalized and dynamic environment. In addition, the use of Large Language Models from different providers and technological environments is a critical factor, given the need for massive computational in frastructure and high energy consumption, which make their adoption unfeasible in the Brazilian context. Therefore, this project proposes to optimize the ontology alignment process, starting from an SBERT + K-Means + Weka baseline, through Deep Learning, with emphasis on continual learning and knowledge distillation techniques, as well as to develop a Green AI framework capable of preserving high levels of Precision, Recall, and F1-Score on the OAEI Anatomytrack, while significantly reducing energy consumption and enabling efficient local execution.