Glossary
Trainable Classifier
A Microsoft Purview ML model trained on examples to identify a category of content like contracts, source code, or healthcare records.
A trainable classifier in Microsoft Purview is a machine-learning model that categorises content based on examples — agreements, source code, financial statements, healthcare records, employee records, legal documents, customer communications, business plans. Microsoft ships pre-trained classifiers for many common categories; organisations can also train custom classifiers by providing 50–500 positive and negative example documents. Used as input to DLP policies, sensitivity-label auto-application, retention auto-application, and insider risk policies — classifiers complement pattern-based sensitive information types (SITs) by detecting content that doesn't fit regex patterns. Requires Microsoft 365 E5 or equivalent Purview licensing.