What is Named Entity Recognition?
You'll learn: What Named Entity Recognition (NER) is and how it identifies and classifies important entities in text.
The Core Idea
Named Entity Recognition is a fundamental NLP task that automatically locates and categorizes specific pieces of information in text. Think of it as highlighting and labeling the "who," "where," and "what" in a document—identifying persons, organizations, locations, dates, monetary values, and other named entities.
How NER Works
NER is formulated as a sequence labeling problem. Instead of classifying an entire sentence (like in text classification), NER assigns a label to each token in the input sequence.
For example, given the sentence:
"Apple Inc. was founded by Steve Jobs in Cupertino."
An NER system would label:
- "Apple Inc." → ORGANIZATION
- "Steve Jobs" → PERSON
- "Cupertino" → LOCATION
Why This Matters
NER is essential for extracting structured information from unstructured text. It powers applications like:
- Information extraction from news articles or documents
- Question answering systems that need to identify entities mentioned
- Search engines that understand what types of entities users are looking for
- Knowledge graph construction by identifying entities and their relationships
The Sequence Labeling Approach
Unlike text classification where you predict one label per document, NER predicts one label per token. Modern approaches use transformer models (like BERT) that have been fine-tuned to assign entity tags to each word or subword token in context, understanding that "Apple" in "I ate an apple" is different from "Apple released a new iPhone."
Key Takeaway: Named Entity Recognition is a sequence labeling task that identifies and classifies named entities (persons, organizations, locations, etc.) in text by assigning labels to individual tokens.