Systems and methods for generating alerts based on potential health conflicts are described. The described systems and methods may provide for identifying unstructured text from a physical environment, performing natural language processing on the unstructured text to produce annotated text, comparing the annotated text to patient medical information to find a potential conflict, and generating an alert based on the potential conflict.