Fact extraction: structured data from unstructured text
- Last Updated: November 20, 2026
- 1 minute read
- Semaphore
- Version 5.12
- Documentation
Beyond tagging documents with concepts, Semaphore can extract structured facts from unstructured content. This includes:
Named entities: People, organizations, locations
Quantities: Dates, monetary values, percentages
Relationships: "Company A acquired Company B for $X"
Fact extraction is powered by:
Natural Language Processing (NLP): Tokenization, part-of-speech tagging, and pattern recognition
Semantic rules: Context-aware logic that understands how entities relate
Use Cases:
Extracting parties and amounts from contracts
Identifying adverse events in clinical trial reports
Capturing customer names and complaint types from support tickets