Presented November 27, 2018.
Unstructured data such as chief complaints and provider notes are an important component of effective Health surveillance. Applying machine learning (ML) and natural language processing (NLP) to unstructured data can often improve surveillance performance over traditional keyword search methods. This presentation explains how health surveillance practitioners can begin applying basic ML and NLP methods to unstructured data, and provides common sense guidelines for which methods provide the greatest improvement for a given level of effort. Lastly the presentation provides an overview of how more involved approaches such as Deep Learning Neural Networks frequently offer even greater performance gains.
Presenter
Drew Levin, PhD, Technical Staff, Sandia National Laboratories