QuickSight Launches Unified Semantic Layer for Multi-Dataset Topics

AWS ML Blog · rss · 2026-07-08

Amazon QuickSight introduces multi-dataset Topics (Public Preview), allowing up to 12 datasets to be added and related within a single Topic. This feature maintains normalized data while upholding centralized governance. **How it Works:** 1. **Data Source & Enrichment**: Supports the SPICE engine and direct queries (e.g., Redshift, Snowflake). Semantic metadata like column descriptions and synonyms can be added independently. 2. **Multi-Dataset Topic**: Acts as a unified container, mapping connection keys between datasets via a JSON file upload, and supports custom instructions for specific business terms. 3. **AI Consumption & Parsing**: When users ask questions, the NLQ engine parses intent, traverses the relationship graph for the shortest connection path, and auto-generates SQL containing JOIN and GROUP BY clauses to return visual results. This pattern allows business users to perform complex cross-dataset analysis via natural language without needing to understand the underlying database schema.

Original post →

More from Apps

Apps channel →