Discover data using semantic search

This document describes how to use project-based semantic search capabilities offered by Dataplex Universal Catalog Search.

Semantic search, powered by Gemini, simplifies data discovery without the need for complex search syntax. It supports natural language queries, so you can search for resources using everyday language. This enhances accessibility for users with varying levels of technical expertise.

Similar to keyword search, semantic search emphasizes the discovery of resources by analyzing the metadata associated with the resources within an organization. Semantic search relies on technical metadata for data discovery. It also supports user-defined metadata, such as aspects. Semantic search focuses on enhancing recall rather than precision.

Pricing

Semantic search in Dataplex Universal Catalog is offered at no charge during Preview.

During Preview, share your feedback and suggestions with the Dataplex Universal Catalog team using the email address dataplex-semantic-search-feedback@google.com.

Required roles

When you search for resources in Dataplex Universal Catalog using natural language, Dataplex Universal Catalog automatically and seamlessly applies the same permissions you have for keyword search.

Console

  1. In the Google Cloud console, go to the Dataplex Universal Catalog Search page.

    Go to Search

  2. To turn on semantic search, click the Query in natural language toggle.

  3. In the search field, enter your query in natural language. The following are some sample queries:

    • Show me the datasets that contain taxi information
    • Find data on vaccine distribution across different countries
    • Get tables with historical temperature data for major US cities
    • Search for hurricane tracking and storm activity datasets
    • US population data by state

  4. Review the search results.

  5. Optional: To view the details of any entry in the search results, click the entry.

REST

To search for resources, use the searchEntries method.

What's next