Evaluating performance of natural language processing for detection of symptoms presence and duration within primary care setting in Singapore

摘要

Purpose: Free-text data in Electronic Medical Records (EMR) are potentially useful to outbreak surveillance but needs prior processing into signs, symptoms and syndromes by natural language processing (NLP) algorithms. We aimed to programmatically read EMR and extract data on signs, symptoms, and symptom duration, and hereby describe the performance of an NLP we designed.