Amazon Quick Sight has introduced a new feature that allows analysts to generate SQL through natural language, enabling complex multi-dataset queries without the need for pre-joining data. This advancement lets users interact with multiple datasets through a chat interface, making it easier to explore data and derive insights. The feature supports up to 12 datasets and allows for flexible joins like outer joins and cross-grain comparisons, expanding the possibilities for data analysis. The change marks a shift in how analytics teams approach data integration and query generation, offering a more dynamic and intuitive experience.
The new functionality, called Chat-powered, AI-generated SQL, enables users to define a semantic layer that includes dataset-level custom instructions, topic-level instructions, field synonyms, and field descriptions. The AI uses this context to generate context-aware SQL at query time, allowing for outer joins, unions, subqueries, self-joins, cross-grain comparisons, and conditional join logic without structural constraints. This approach differs from the traditional method of defining explicit relationships, which requires a data engineer to pre-join datasets before any analysis can occur. The AI-driven method is designed for exploratory analytics, ad-hoc questions, and power users who need flexibility in their data exploration.
The source explains that Amazon Quick Sight’s Multi-Dataset Topics now support two modes: one where relationships are defined explicitly, and another where the AI generates SQL based on semantic guidance. The latter approach uses a layered framework called the Semantic Guidance Stack, which includes seven layers of metadata that guide the AI in generating accurate SQL. These layers include dataset-level instructions, topic-level instructions, field synonyms, field descriptions, column exclusions, and calculated fields. The more precisely these layers are populated, the more accurate the generated SQL and the more reliable the results.
Source: awsml
