Ontology Information-augmented Human Phenotype Concept Recognition

Automatic phenotype concept recognition from text is of great significance for disease analysis. Existing ontology-driven phenotype concept recognition methods mainly utilize concepts and synonyms information within the ontology, without fully considering the rich information of the ontology. To address this issue, this paper proposes an ontology-enhanced method for human phenotype concept recognition, which utilizes advanced large language models for data augmentation. We design a deep learning model enhanced with ontology vectors to improve concept recognition performance. Experimental results on the GSC+ and ID-68 datasets demonstrate that the proposed method achieves state-of-the-art results.