Entity Linking

Definition

Entity Linking is the process of connecting an entity mention in text to a specific canonical entity in a knowledge base, knowledge graph or other reference system.

Why it matters

Entity linking allows AI systems to move from recognising a name to understanding which real-world entity that name refers to. This is important when different entities share similar names or when the same entity appears under multiple surface forms.

How it works

Mention identification

  • An entity mention is detected in text, often using Named Entity Recognition or related methods.

Candidate generation

  • Possible canonical entities are identified from a knowledge source.
  • Candidates can be generated using names, aliases, context or other entity signals.

Candidate selection

  • The system evaluates context and selects the most plausible canonical entity.
  • If evidence is insufficient, the link may remain unresolved.

How Netsleek uses the term

Netsleek uses Entity Linking to distinguish the act of connecting a mention to a canonical entity from broader Entity Resolution and Entity Disambiguation. This is especially relevant to brand discoverability where names, products and parent companies can be confused.

Comparisons

How Entity Linking Connects Across AI Systems

Entity Linking does not operate in isolation. It connects to other established concepts across the AI search and retrieval process:

  • Uses canonical entities: Canonical Entity records provide the target identity for linking.
  • Uses context: Context Resolution and semantic signals can help choose between ambiguous candidates.
  • Connects to knowledge systems: Knowledge Bases and Knowledge Graphs commonly provide canonical entity identifiers.
  • Supports retrieval: Linked entities can improve entity-aware search and retrieval.

Related glossary concepts

Common misinterpretations

  • Entity linking is not the same as simply matching a name string.
  • A detected brand name can still be linked to the wrong entity.
  • Entity linking does not require that every entity be stored in a public knowledge graph.

Summary

Entity Linking maps a textual entity mention to a canonical real-world entity. It connects language processing with knowledge systems and helps AI systems distinguish which person, brand, product or organisation a mention refers to.