Slimmer Banner 28 05 2021

Doing almost as much with much less: A Case Study in Biomedical Named Entity Recognition- Dr. Stéphan Tulkens

Can a simple string-matching method compete against a big supervised model?


Our Senior Machine Learning Engineer, Dr. Stéphan Tulkens published an article about biomedical named entity recognition with knowledge bases. In a previous post, Sybren Jansen, our Head of AI, showed that biomedical named entity recognition is often tackled using very big models. The use of large models incurs significant cost in training and inference time. In this article, he shows that a very simple string matching metric, coupled with some smart optimizations, comes a long way in terms of performance, in only a fraction of the time. Read more here.