Q&A Natural Language Queries: Linguistic Schema Training

Business executives frequently want answers to rapid, unexpected questions without waiting days for a BI team to build a new dashboard:

• "Show total sales by state as a map"
• "Top 5 customers by profit margin in 2025"
• "What were laptop sales last quarter?"

The Q&A Visual uses natural language processing (NLP) to parse conversational human questions into dynamic DAX queries and auto-generate the correct visual on the fly!

1. How the Q&A Engine Works

2. Q&A Tooling & Linguistic Modeling

In real organizations, business users do not always use the technical column names present in the database. A sales rep might say "turnover", "income", or "top-line" when referring to the column Fact_Sales[SalesAmount].

Training the Q&A Engine:

In Power BI Desktop, go to Modeling $\to$ Q&A Setup:

1. 1
   Synonyms: Add business aliases to tables and columns:

• Dim_Customer $\to$ Synonyms: Client, Buyer, Account, Purchaser.
• Fact_Sales[Profit] $\to$ Synonyms: Earnings, Net Return, Margin.

1. 1
   Teach Q&A: Train the engine on industry-specific phrases:

• Define: "A 'VIP Customer' is a customer whose Total Sales > $50,000".

1. 1
   Review Questions: Inspect what real users are typing in the Power BI Service and fix un-recognized terms.