Text Analytics for Health is a capability within Microsoft's Azure Language service that reads unstructured medical text and labels the clinical information in it. It uses machine learning to identify entities in doctor's notes, clinical documents and electronic health records, such as diagnoses, medication names, symptoms and age. A single API call performs four functions: named entity recognition, relation extraction, entity linking and assertion detection. A newer addition extracts social determinants of health and ethnicity mentions. It is available as a hosted API or as an on-premises Docker container.
This service is for developers, data engineers and health IT teams at hospitals, insurers, health software companies and research groups who need structured data out of free-text clinical records. Clinical informatics specialists use it to map mentions to standard vocabularies, while product teams embed it in applications that analyze patient documentation. Microsoft states it is provided as is and is not a medical device, diagnostic tool or substitute for professional judgment, so decisions affecting individuals should have human oversight. Customers must also separately license the source vocabularies they plan to use.
Typical scenarios include pulling key data from discharge summaries to support healthcare analysis, structuring physician notes so they can be searched, and linking extracted concepts to medical vocabularies for coding and cohort selection. Researchers can scan large document collections for medications, conditions or symptoms, and assertion detection helps tell whether a condition is affirmed, negated or uncertain. Providers can surface social determinants such as substance use mentions to understand patient context. Teams can try it in the Microsoft Foundry portal with an existing Language resource, or follow the quickstart and how-to guides to make a first request.
