Abstract This paper describes our participation in Track 3 of the BioCreative VIII shared task focused on extracting and normalising genetic phenotypes from dysmorphology physical examination reports. We focus on disjoint entity spans which make up around 14% of the mentions. We developed an approach, DiscHPO, that extracts and normalises both continuous and discontinuous spans. The system consists of two components: a sequence-to-sequence named entity recognition model and an entity normaliser based on a Sentence-Transformer and a Cross-Encoder re-ranker. The best performing model for entity normalisation obtained an F1 score of 0.7229 on the test data, whilst the best model for span extraction achieved an F1 score of 0.6647. This article is part of the Proceedings of the BioCreative VIII Challenge and Workshop: Curation and Evaluation in the era of Generative Models.