Entity Linking
Entity Linking is the process of mapping text mentions of entities to unique entries in a knowledge base (e.g., Wikidata).
Entity Linking maps text mentions to unique knowledge base entries – resolving ambiguities and connecting NLP with Knowledge Graphs.
Explanation
Entity Linking resolves ambiguities: "Apple" can be the company, the fruit, or a music label. The algorithm uses context to find the correct entity in the KB.
Marketing Relevance
Entity Linking improves SEO through semantic annotation, enables more precise content recommendations, and bridges NLP and Knowledge Graphs.
Example
A news aggregator uses Entity Linking to automatically connect articles with Wikipedia entities and cluster them by topic.
Common Pitfalls
Missing entities in the KB (NIL problem), incorrect disambiguation for rare entities, language/domain bias of training data.
Origin & History
Entity Linking evolved from Named Entity Recognition (MUC-6, 1995). The TAC-KBP competition (2009) established standardized evaluation. Modern systems like BLINK (Facebook, 2019) use dense retrieval for real-time linking.
Comparisons & Differences
Entity Linking vs. Named Entity Recognition (NER)
NER identifies and classifies entities in text (person, location, org); Entity Linking additionally maps them to a knowledge base.
Entity Linking vs. Coreference Resolution
Coreference Resolution links different text mentions of the same entity ("he", "the CEO"); Entity Linking maps to an external KB.