Home › AI / ML › NLP › Named Entity Recognition (NER)

Named Entity Recognition (NER)

NLP information extraction entities sequence labeling

Pulling the facts out of text

Named Entity Recognition finds the real-world things a sentence mentions — people, companies, places, dates, amounts — and labels each with its type.

It turns free text into structured data. From "Apple acquired Beats for $3B in 2014", NER extracts Apple (organization), Beats (organization), $3B (money), 2014 (date) — facts a database can store and query.

Spot the entities

Watch a sentence get scanned, each entity highlighted and tagged with its category.

Typical entity types

PERSON people

Tim Cook, Marie Curie.

ORG organizations

Apple, the UN, Stanford.

GPE / LOC places

Paris, Japan, the Nile.

DATE / TIME when

2014, last Tuesday, noon.

MONEY / PERCENT amounts

$3B, 25%.

PRODUCT / EVENT things

iPhone, the World Cup.

Where it's used

Search & KGs structure

Build knowledge graphs and richer search from raw articles.

Resume / document parsing extraction

Pull names, companies, dates from CVs, contracts, invoices.

Redaction privacy

Find and mask personal information automatically.

How it works

NER is sequence labelling, often in BIO format (Begin/Inside/Outside an entity). Modern systems fine-tune a transformer; classic ones use CRFs. spaCy ships strong NER out of the box, building on POS tags.