5.1.2. Entity Recognition
Entity recognition identifies and categorizes named entities in text—specific things like people, places, organizations, dates, and quantities. Think of it like a highlighter that automatically marks different types of information in different colors: people in yellow, places in blue, dates in green.
Key characteristics:
- Input: Text containing mentions of real-world entities
- Output: Entity names with category labels and positions in text
- Categories include: Person, Location, Organization, DateTime, Quantity, URL, Email, Phone
How it differs from key phrase extraction:
- Key phrases: "machine learning", "cloud computing" (general concepts)
- Entities: "Microsoft" (Organization), "Seattle" (Location), "January 2024" (DateTime)
Entity Linking goes further by connecting recognized entities to a knowledge base (like Wikipedia), providing additional context and disambiguation. For example, "Apple" could mean the fruit or the company—entity linking determines which based on context.
PII Detection is specialized entity recognition for personally identifiable information:
- Names, addresses, phone numbers
- Social Security numbers, credit card numbers
- Email addresses, IP addresses
- Medical record numbers
Common scenarios:
- Extracting company names from news articles for market intelligence
- Identifying people mentioned in legal documents
- Pulling dates and deadlines from contracts
- Detecting and redacting PII for privacy compliance
- Building knowledge graphs from unstructured text
⚠️ Exam Tip: PII detection is a specialized form of entity recognition focused on sensitive personal data. Use it for compliance scenarios involving GDPR, HIPAA, or data privacy requirements.
