About AI Descriptions for Datasets

AI Descriptions are semantic metadata that explain a column’s business meaning and analytical role. Oracle Analytics AI Agents use this information to identify the appropriate datasets areas for analytical and research questions and to generate accurate SQL statements when answering those questions.

For example, an AI Description can explain what a column represents, how it is used to filter or group data, or how it relates to other columns.

To learn how to add AI Descriptions to a dataset, see Add AI Descriptions to a Dataset.

Why Add AI Descriptions to a Dataset

AI Descriptions are semantic metadata that provide Oracle Analytics AI Assistant with additional business context about a dataset and its columns. The Assistant uses an AI Description when a user’s question is semantically related to that dataset or column.

Use AI Descriptions to explain information that a dataset name, column name, or synonym alone doesn’t convey. Well-written descriptions can improve the accuracy and consistency of Assistant responses by helping the Assistant select the appropriate column, apply filters correctly, and interpret values as intended.

When to Add an AI Description

Add an AI Description when users or the Assistant need to understand an important business rule or detail to interpret the data correctly. For example, use an AI Description to:

• Distinguish between similar columns, such as Invoice Forecast and Opportunity Forecast.
• Explain a measure’s grain, unit, or scale.
• Describe how column values are formatted.
• Explain a relationship or dependency between columns.
• Identify important inclusions or exclusions.

For example, if a measure is stored in thousands of dollars, describe that unit explicitly. If a Consolidated Total value already includes other invoice types, describe the rule so the Assistant doesn’t aggregate the total with its component values.

AI Descriptions, Synonyms, and Agent Instructions

Column names and synonyms help the Assistant identify a column. AI Descriptions explain the business meaning of the dataset or column and how its values should be interpreted.

AI Descriptions don’t replace Agent supplemental instructions. Supplemental instructions define how an Agent behaves and responds to users, whereas AI Descriptions define the semantics of the dataset. A dataset or column AI Description can be used by multiple Agents.

Guidelines for Writing AI Descriptions

Write concise, specific descriptions that state the business context the Assistant needs. Include the details that affect query logic or result interpretation, such as units, value formats, and dependencies.

Don’t repeat the column name or use an AI Description only to list alternate names. Add synonyms for alternate names, and use an AI Description for business meaning that names and synonyms can’t express.

You can select Generate all to create suggested AI Descriptions. Review generated descriptions before saving them, and retain or edit only descriptions that add accurate, meaningful context for your data.