Entity Linking.prompt

# Entity Linking and Disambiguation

## Overview

Entity linking is the process of identifying mentions of entities in text and linking them to canonical entries in a knowledge base (Wikipedia, Wikidata, domain-specific databases). This resolves the ambiguity inherent in natural language — "Apple" could refer to the fruit or the company, "Paris" to the city in France or Texas. LLM-based entity linking combines contextual understanding with world knowledge to make accurate disambiguation decisions.

Production entity linking systems are critical for knowledge graph population, semantic search, content recommendation, and RAG systems that need to ground text references in structured data. Modern approaches use LLMs both for mention detection and for disambiguation by leveraging the model's parametric knowledge of entities.

## Key Concepts

- **Mention detection**: Identify spans of text that refer to entities
- **Candidate generation**: For each mention, generate a list of possible entity candidates from the knowledge base
- **Disambiguation**: Use context to select the correct entity from candidates
- **NIL detection**: Identify mentions that don't have a corresponding entity in the knowledge base
- **Cross-document coreference**: Link entity mentions across multiple documents
- **Entity typing**: Classify entities by type (Person, Organization, Location, etc.)
- **Confidence scoring**: Assign probability scores to linking decisions

## Implementation Patterns