Hi! The upcoming version of Prodigy will include a new UI for annotating overlapping spans. You can read more about it here: ✨ Prodigy nightly: spaCy v3 support, UI for overlapping spans, improved feeds & more
However, just for completeness (and others coming across this thread later): the data you'd create here wouldn't be suitable for a regular named entity recognition model. Those implementations typically assign token-based tags, and by defintion, every token can only be part of one entity (which is also one of the reasons NER generally works so well). So if your goal is to train a model to predict those annotations, you'd want to choose a different implementation, e.g. one that predicts labels for arbitrary spans of text.