Automated and Explainable Deep Learning for Clinical Language Understanding

Explore advanced deep learning techniques for clinical language understanding in this 35-minute conference talk from Databricks. Dive into the application of natural language processing (NLP) in healthcare AI, focusing on unstructured medical notes. Learn about the deep learning techniques, explainability features, and NLP pipeline architecture used to tackle healthcare-specific challenges. Discover how Spark NLP for Healthcare, BERT embeddings, and healthcare-specific embeddings are leveraged to understand clinical terminology and extract specialty-specific information. Gain insights into the use of transfer learning to minimize task-specific annotation requirements and the integration of MLflow with Spark NLP for experiment tracking. Explore the automated deep learning system's capabilities in training, tuning, and measuring models based on clinical annotations. Understand the annotation process, guidelines, and the importance of automation in handling diverse clinical language across providers, document types, and geographies. Learn how explainable results are achieved by highlighting evidence in text for extracted semantic facts. This talk is ideal for data scientists, software engineers, architects, and leaders designing real-world clinical AI applications, offering valuable lessons in applying cutting-edge NLP and deep learning techniques in healthcare.

Introduction
About Roche
Decision Support Portfolio
Search for a Cancer Patient
Pathology Report
What is NLP
Background
Optical Character Recognition
Entity Resolution
The Process
The Workflow
The Pipeline
Spark NLP
Spartan NLP
Accuracy
Scalability
Optimized Hardware
Clinical Entity Recognition
Clinical Entity Resolution