Map the physical layer and semantic layer through semantic mapping (in preview)

Important This topic is applicable only to the latest UI.

If you have defined the semantic layer and want to map the physical layer to it, you get help from the semantic mapping feature.
You can also manually link the Column assets and Data Attribute assets. For information on that option, go to Manually map the physical layer and semantic layer (in preview).

About semantic mapping in the semantic assistant (in preview)

Semantic mapping is a Collibra AI feature that helps data stewards map the semantic layer to other assets in Collibra. With semantic mapping, you don't need to manually map the assets. The tool suggests Data Attribute assets for Column assets and creates relations when suggestions are accepted. These suggestions are based on defined criteria such as descriptions and selected data attributes. The mapping between the semantic and physical layers adds context, ensures consistency, and simplifies analysis by AI models.

Image of the flow from the start of the automation process to the generated Data Attribute suggestions.

  • Once you have defined the Data Model with Data Entity and Data Attribute assets, you can start the semantic mapping process for a table or view (1).
  • Semantic mapping guides you through a series of options to define the criteria that Collibra AI must take into account (2).
  • The process collects the data (3) and sends a request to Collibra AI (4).
  • When suggestions are available (5), they are shown (6).
    You can then verify the suggestions and either accept or reject them. If you accept a suggestion, a "Data Attribute represents / is represented by Column" relation is created between the Column asset and the Data Attribute asset.

Important 
Currently, semantic mapping works only when the table has less than 250 columns. If the table contains more than 250 columns, the mapping fails and no mapping results are available.

Start the automated mapping process (in preview)

Prerequisites

Steps

  1. On the main toolbar, click , then Stewardship.
  2. In the submenu, click Semantic layer.
  3. Click Semantic Assistant.

    A page opens showing a data explorer on the left-hand side. The panel shows all available data sources in Collibra.

    Image of the Semantic Assistant page

  4. In the data explorer, complete the following steps to find the columns you want to link to Data Attribute assets.
    1. Expand the database and then the schema to show the available tables and views.
    2. Click the table or view to show the columns on the right-hand side.

    Image of a selected Table asset in the data explorer on the left and the related Columns on the right

  5. Click Map data attributes.
    The Map data attributes dialog box appears.
    By default, Collibra AI uses the descriptions of the columns and the related tables, views, schemas, and databases, and the accepted data classifications of the columns to create Data Attribute suggestions.
  6. Define semantic model targets to limit the attributes from which Collibra AI can choose to make suggestions.
    • Select at least one target: a data model, one or more data entities, or both.
    • Use the Target data model drop-down list to filter the Target data entities list if needed.
    • If you select only a data model, all attributes for all its data entities are considered.
    • If you select a data model and specific data entities from that model, only attributes for those data entities are considered.
    • If you select only the data entities, the attributes for those data entities are considered.

    Image showing the operating model: Data Model contains Data Entity, and Data Entity contains Data Attribute

  7. Click Map.

    The mapping process starts. Collibra AI analyzes the selected options and tries to make a suggestion for all columns in the table or view that aren't yet mapped to a Data Attribute.

    • If a Data Attribute is found, the suggestion is shown.

      • You can follow up on the progress in the Activities page.
      • Only one suggestion is provided per column.
      • The full name of the Data Attribute asset is shown as a link. You can click it to open the Data Attribute asset page. This full name can provide additional context when you verify the suggestions.
      • The suggestion is marked with the Collibra AI icon.
      • A similarity score, shown as a percentage, indicates how closely the Data Attribute information matches the column information analyzed by Collibra AI. This score can assist in verifying the suggestions.

    • If no Data Attribute is found, no suggestion is shown.

    Important 
    Currently, semantic mapping works only when the table has less than 250 columns. If the table contains more than 250 columns, the mapping fails and no mapping results are available.

  8. Verify the suggested Data Attributes.
    You can filter the visible Column assets by similarity score using the Data Attribute column. In the filter dialog box, use the Similarity score slider to show only suggestions with a score above a specific value.

    Image of where to click the open the filter dialog box

    Image that shows how the Columns assets are filtered based on the similarity score slider value

    You can then accept or reject the Data Attribute suggestions.
    The Data Attribute asset and the Column asset aren't linked until you accept the Data Attribute suggestion.

Accept Data Attribute suggestions (in preview)

You can accept suggested Data Attributes in the following ways.

  • In the Data Attribute suggestion, click the Check icon.
    Image of the Check icon
  • Select the checkbox for the columns for which you want to accept the suggestion and click Accept Data Attribute.
    Use this method to accept suggestions for multiple columns in a single action.

When you accept a Data Attribute suggestion for a column:

  • The similarity score and the Collibra AI icon are removed.
  • The display name of the Data Attribute asset is shown.
  • A Data Attribute represents / represented by Column relation is created between the Data Attribute asset and the Column asset. This relation has publicID DataAttributeRepresentsColumn.

Reject Data Attribute suggestions (in preview)

You can reject suggested Data Attributes in the following ways.

  • In the Data Attribute suggestion, click the x icon.
  • Select the checkbox in front of the Data Attribute that you want to reject and click Reject Data Attribute.
    Use this method to reject suggestions for multiple columns in a single action.

When you reject a Data Attribute suggestion for a column, the suggestion is removed.

Tip When you ran the semantic mapping, and you want to manually map a column afterward, the same target data model and entities as used during the automatic mapping are already selected in the dialog box. If both a data model and data entities were selected, only the data entities remain selected because the data model acts only as a filter in this case.

What's next

You can see the connected assets for Tables and Columns through the Semantic layer widget in the asset page and in the Semantic Models overview.

Related topics