Understanding how watsonx BI generates a response

When you ask watsonx BI a question, it analyzes and interprets your question, identifies relevant data sources, and determines the best way to respond to the question (a text response or visualization).

Watsonx BI uses your input and the data that you have access to in the form of metrics and the enriched metadata, to understand the business logic. It then uses column titles, descriptions, samples values, and SQL examples, and instructions to find the most relevant data to generate a context-aware response.

Response generation process

Interpretation and analysis - watsonx BI interprets your question
Context rephrasing - watsonx BI might rephrase the question based on the historical context of previous questions
Key word extraction - watsonx BI pulls out key search terms from the question
SQL generation - watsonx BI uses the following to convert your question into SQL:
- Rephrased question and key terms
- Column labels, column descriptions, sample values in metrics, and enriched metadata
- Asset labels and asset descriptions
- General instructions
- SQL examples
Response generation - watsonx BI uses queried data, the rephrased question, and final phrasing instructions to generate a text response or visualization