Maternal health outcomes are essential indicators of overall health care quality and societal well-being. However, in the United States, the maternal health surveillance is often inaccurate, restricting the clinical utility of the data gathered. The limits imposed by these inaccuracies restrict timely policy responses and hinder effective innovations, despite the increasing availability of electronic health records. This paper explores the potential use of natural language processing in improving maternal health surveillance. By combining rule-based linguistic processing with machine learning, natural language processing can transform narrative text into structured, analyzable data, allowing it to be used for predictive purposes, as well as the development of real-time public health surveillance systems.