Deep Learning in Anatomy Ontology Matching: The VIEWAnatomy-DL System at OAEI 2026

This paper presents VIEWAnatomy-DL, an interactive ontology alignment system evaluated in the Anatomy track of OAEI 2026. The system combines lexical similarities, K-Means, Sentence-BERT em-beddings loaded via the Deep Java Library (DJL), a hybrid candidate generator, an SBERT-aware clustering threshold, a global bidirectional 1:1 filter and a final similarity floor. It reaches precision 0.9509, recall 0.7672 and F-measure 0.8492 (1,223 correspondences). Among the 14 official systems plus VIEWAnatomy-DL (15 entries), it ranks 7th in precision and 11th in F-measure. The comparison shows that generic SBERT embeddings are an effective low-cost step into deep learning, while domain-specific knowledge (SapBERT, UMLS) and LLM oracles remain the key differentiators of top systems. A roadmap targeting F-measure in the 0.955-0.960 range for OAEI 2027 is proposed.