Build a unified semantic layer across datasets with multi-dataset Topics in Amazon Quick

This article announces multi-dataset Topics in Amazon QuickSight, a feature that enables building unified semantic layers across multiple normalized datasets with AI-powered natural language querying.

• Combine up to 12 datasets in a single topic with explicitly defined relationships and join keys
• NLQ engine automatically traverses relationships and constructs appropriate SQL joins based on user intent
• Support for SPICE datasets and Direct Query against Redshift, Athena, S3 Tables, Snowflake, and Databricks
• Enrich datasets with semantic metadata including descriptions, synonyms, semantic types, and calculated fields
• Define relationships using JSON configuration files mapping join keys between dataset pairs
• Add custom instructions for domain-specific terminology, date logic, and business definitions
• Use topics in analysis sheets and chat interface for cross-dataset natural language questions
• Permissions model supports Owner and Viewer roles with row-level and column-level security enforcement

Multi-dataset Topics preserve normalized data models, reduce duplication, and provide business users with richer cross-domain answers through a governed semantic layer.