Ai+ Training

Topic#1 "State of the art NLP in Healthcare"

Many critical facts required by healthcare AI applications like patient risk prediction, cohort selection, and clinical decision support are locked in unstructured free-text data. Recent advances in deep learning and transfer learning have raised the bar on achievable accuracy for tasks like named entity recognition, assertion status detection, entity linking, de-identification, and relation extraction, using novel healthcare-specific networks and models. 

This session introduces Spark NLP for Healthcare:  the world’s most widely used healthcare NLP library, which holds a majority of top spots on accuracy leaderboards, is optimized to run on modern CPU's & GPU’s and is natively able to scale to run training & inference on Spark clusters. Live & freely available Python notebooks will be used to demonstrate how to achieve state-of-the-art accuracy & speed on common medical NLP tasks.

Topic#2: Working with German Radiology Reports

In this talk, Katrin will cover various challenges one encounters when working with natural language texts in form of radiology reports and potential solutions for them. In addition to general challenges like different styles across different hospitals, speaker will also cover language-specific challenges for German written in Switzerland. German used in Switzerland is slightly different from the one used in Germany, the most prominent would be that in Switzerland we do not use ß but write ss instead. 

Working in healthcare requires us to put exceptional effort into making our predictions more accurate. For Katrin, this meant addressing each of these challenges individually whereby they often proved to be more complicated than first expected. For example, simply replacing ß with ss in the data set to normalize all the input data doesn’t do the trick.

Local ODSC chapter in Zurich, Switzerland

Co-organized with NLP Zurich community