Using NER and LLMs to enhance de-identification of semi-structured EMR data – preliminary results and lessons learned
Today I presented a poster at the GMDS 2023 congress in Heilbronn Germany. The poster was about using Named Entity Recognition (NER) and Language Models (LLMs) to enhance the de-identification of semi-structured Electronic Medical Record (EMR) data. In essence we used open-source LLMs to augment our de-identification pippeline for semi-structured EMR data. The LLMs were used to identify entities that were not covered by the NER model we used. Since all of this had to be done on-site (on-premise) and with limited resources, you can probably imagine that the results were’nt stellar....