Named Entity Recognition (NER)

Definition

Named Entity Recognition (NER) is an NLP task that identifies spans of text referring to named entities and classifies them into categories such as organisations, people, locations, products or other entity types.

Why it matters

NER is a foundational step in many entity-aware systems. Before a system can link, disambiguate or reason about an entity, it often needs to detect that an entity mention exists in the text.

How it works

Mention detection

  • The system identifies the words or token spans that refer to an entity.
  • For example, a company name, person name, location or product can be detected as an entity mention.

Entity typing

  • Detected mentions can be assigned an entity type such as organisation, person or location.
  • The supported entity classes depend on the model or application.

Downstream processing

How Netsleek uses the term

Netsleek uses NER as the established upstream entity-processing mechanism that helps explain how brand and product mentions can be detected before being linked to canonical entities. It is not treated as evidence that a system has correctly resolved the entity.

Comparisons

  • NER vs Entity Linking: NER detects an entity mention. Entity Linking connects that mention to a canonical entity.
  • NER vs Entity Disambiguation: NER identifies the mention. Disambiguation selects the correct entity when multiple candidates are possible.
  • NER vs Entity Resolution: Resolution determines whether records or references correspond to the same real-world entity.

How Named Entity Recognition Connects Across AI Systems

Named Entity Recognition (NER) does not operate in isolation. It connects to other established concepts across the AI search and retrieval process:

  • Part of NLP: NER is a standard Natural Language Processing task.
  • Feeds Entity Linking: Detected mentions can be mapped to canonical entities or knowledge bases.
  • Supports entity-aware retrieval: Recognised entities can be used as signals in Contextual Entity Search or other retrieval systems.
  • Supports knowledge graphs: Recognised and linked entities can connect text to graph-based knowledge.

Related glossary concepts

Common misinterpretations

  • NER does not prove that the correct real-world entity has been identified.
  • Entity detection and entity linking are separate tasks.
  • NER categories vary across models and applications.

Summary

Named Entity Recognition detects and categorises entity mentions in text. It is an upstream NLP mechanism that often precedes entity linking, disambiguation and entity-aware retrieval.