MedLEE, a Natural Language Processing Service, on the Public Health Research Grid

Much clinical information needed for biosurveillance and monitoring occurs in the form of highly telegraphic unstructured text, which is not suitable for accurate retrieval. A natural language processing (NLP) system aims to process vast amounts of text in a reasonable amount of time, and to accurately extract and encode the relevant biomedical information so that it is in a standard format, thereby rendering the text into a more computable form for further processing and accurate retrieval. However, NLP methods are complex, difficult, and expensive to implement, and many focus on extracting highly specialized information only. MedLEE is a comprehensive natural language processing system for clinical documents, which has been shown to perform well for numerous applications. We will describe the efforts and challenges involved in making MedLEE available as a service on the PHGrid, and will demonstrate MedLEE on the research grid.