Named Entity Recognition

Named Entity Recognition

Named Entity Recognition

Named entity recognition locates spans of text referring to defined categories and labels them. General systems recognize persons, organizations, locations, dates, and monetary amounts; legal systems add categories such as courts, statutes, case citations, and medical providers.

It is the first step in most extraction pipelines.

Alternative Names:

NER, Entity Tagging

Why it Matters?

Recognition is only useful when paired with resolution, since identifying four references to a physician does not establish they are the same physician. Legal documents also present recognition problems general systems handle poorly: party designations that shift between name and role, entities referenced only by abbreviation after first mention, and provider names embedded in handwritten notes that OCR imperfectly.

Frequently Confused with

Related terms

Frequently asked questions

What entity types matter in litigation documents?

What entity types matter in litigation documents?

Persons and their roles, organizations, medical providers, dates, monetary amounts, courts, case citations, statutes, and document identifiers such as Bates numbers.

Why is recognition insufficient by itself?

Why is recognition insufficient by itself?

Because it identifies references without determining which refer to the same entity, which requires resolution before the output is usable.