Semantic Models¶
Semantic Models provide a foundation of business context between structured datasets and the LLMs that query them. They help in translating natural language queries into precise, executable SQL.
To create a semantic model, you define Entities (conceptually mapping to datasets/tables) and Attributes (conceptually mapping to columns), how they relate to each other, as well as business metrics (data aggregations) on them.
Semantic models add business meaning to physical data, which provides a consistent business view of your data, and enhance decision making, notably for AI Agents.
Key benefits of DSS Semantic models include:
Agnostic Data Connectivity: Users can create semantic models on top of datasets from any DSS data source
Structured Business Logic: Structure your company-specific semantic concepts through the configuration of entities, attributes, metrics, filters, and relationships.
LLM assisted Model Generation: Accelerate the creation of semantic models through LLM-assisted generation leveraging existing metadata and documentation.
Validation Playground: Test, refine, and verify semantic models built.
Lifecycle Management: Semantic models support versioning, allowing teams to maintain and iterate on different versions while designating a specific one as active.
Agentic Tool for Semantic Model Query: Dataiku Agents can leverage the semantic of semantic models to perform high-quality queries in the underlying data, for highly performant agents