Clinical Entity Linking

In Natural Language Processing (NLP) an common and important task is entity linking is the entails establishing correspondences between free-form text and a structured set of concepts. This is even more challenging in medical or clinical entity linking which involve a large number medical concept unique identifiers (CUIs) as well as domain specific abbreviations and assumptions.

There is an interesting new Driven Data competition, SNOMED CT Entity Linking Challenge at https://www.drivendata.org/competitions/258/competition-snomed-ct/page/816/ with a new dataset based on MIMIC IV (and the corresponding security and required training).

I'm interested in exploring some of the recent LLM and embedding approaches on this data ... anybody want to enter and/or team up?