Named Entity Recognition (NER)

What is Named Entity Recognition (NER)?

Named entity recognition (NER) is a natural language processing (NLP) technique that automatically identifies and classifies named entities with specific, meaningful items within text. In document processing, NER locates and labels entities such as:

• Organization names
• Person names
• Dates and monetary values
• Addresses and product identifiers
• Regulatory references

It extracts these as structured data from unstructured document content.

NER a foundational component IDP

NER is a foundational component of Intelligent Document Processing (IDP). It enables systems to identify not just where a value appears in a document, but what type of entity that value represents.

Why semantic labeling matters

• Distinguishes a date in a "payment due" context from a date in a "contract signed" context
• Recognizes that "Acme Corp." in an invoice header is the supplier entity, not a product description
• Supports more accurate extraction, validation, and downstream integration than positional or pattern-based approaches alone

How advanced NER works in Document AI

Advanced NER in Document AI combines multiple techniques to handle the complexity of real-world business documents:

Core techniques

• Rule-based entity dictionaries: Establish baseline recognition for common, predictable entity patterns
• Machine learning models: Adapt to variability across document types and layouts
• Large language models (LLMs): Understand entity meaning in context, resolving the ambiguity that rigid rule systems cannot