Sybren Jansen, Machine Learning Engineer and Head of AI at Slimmer AI, recently had his article, “Who’s Who and What’s What: Advances in Biomedical Named Entity Recognition (BioNER)”, published in Towards Data Science.

Named Entity Recognition (NER) involves tagging entities in unstructured text using categories such as person, organization, etc. This task is one of the building blocks of NLP and is used in many downstream tasks like question answering, topic modeling, and information retrieval.

Applying NER to the biomedical domain (BioNER) is particularly challenging as data is often not freely available, annotation of data requires expert knowledge, and the space of biomedical concepts is enormous.

In his article, Sybren provides an extensive overview of BioNER research to help tackle these challenges. From techniques on dealing with imperfect datasets, due to annotator disagreement or bias, to transfer, multi-task, and few-shot learning, all the way to knowledge-based systems that do not rely on supervised train data at all, but utilize external knowledge like ontologies.

Given that we’re in the middle of a global pandemic the need for high-performing BioNER systems has never been higher, they’re fundamental in text mining systems to facilitate COVID-19 studies, amongst others.

Follow us on LinkedIn and Twitter for more stories like this.