Embedding-Based Ontology Term Recommendation System for FAIR Data Publishing Workflows

FAIR (Findable, Accessible, Interoperable, and Reusable) data publications are important for enabling open energy research across interdisciplinary domains. The realization of the FAIR principle for data still faces many challenges, such as the diversity of data formats, semantic heterogeneity, lack of formalized ontologies, and error-prone manual annotation of data. These challenges impede the effective sharing and integration of energy data. To address these issues, we propose an automated ontology term recommendation system based on ontology embedding and semantic similarity, aiming to facilitate FAIR data publication in energy research. Our recommendation system utilizes contextual embeddings to automate semantic annotation of heterogeneous energy datasets by linking data elements to predefined energy-specific ontologies and referring to the top-K most relevant ontology concepts. The proposed system is designed to significantly streamline the semantic annotation process for energy researchers, thereby accelerating the process of open energy research. For evaluation, we test our system on the SemTab Challenge 2024 dataset provided by the Ontology Alignment Evaluation Initiative (OAEI). ... mehrThe experimental results show that our system improves the matching accuracy significantly compared to the baseline system