Semantic Data Type
The Semantic Data Type (SDT) specification provides an intelligent bridge between data type concepts and how they are commonly represented in files, databases, and programming environments. By capturing key aspects—including meaning and purpose, usage, formatting, and validation rules—it enables both humans and machines to develop a deep, actionable understanding of the underlying data.
Core Pillars of Semantic Types
Conceptual Alignment
Semantic types are aligned to standard ontologies like SKOS, Wikidata, and LCSH. They represent discrete real-world concepts, not just raw primitive data types.
Physical Storage Mappings
Mappings defined across multiple environments, detailing how a type is stored in databases (SQL, PostgreSQL, BigQuery), represented in programming (Python, Rust, TypeScript), or exchanged (XSD).
Agentic Skills & Instructions
Built-in instructions and system prompt snippets designed for LLMs to interpret, parse, normalize, and reason with variables matching the semantic data type.
Algorithmic Checksums
Multi-layered validation combining structure-matching regular expressions with complex algorithmic check validations (such as Luhn, Modulo-11, and Modulo-97).
Illustrative Examples
Explore how Semantic Data Types elevate primitive data representations into self-documenting, validated, and agent-actionable assets.
US Social Security Number
A sensitive national identifier allocated for taxation and social security administration.
VARCHAR(11) / str
International Bank Account
A standardized global system for identifying financial accounts across national borders.
VARCHAR(34) / str
Canadian Postal Code
A six-character alphanumeric string that forms part of a postal address in Canada.
VARCHAR(7) / str
Explore the Open Specification
The structure and properties of Semantic Data Types are formally modeled in LinkML (Linked Data Modeling Language), providing absolute machine-readability, automated compilation to Python/Rust/TypeScript, and rigorous metadata schema compliance.