Changing the semantic type of a column - Cloud

Talend Cloud Data Preparation User Guide

English (United States)
Talend Cloud
Talend Data Preparation
Administration and Monitoring > Managing connections
Data Quality and Preparation > Cleansing data
Data Quality and Preparation > Managing datasets

When you add a dataset, the application automatically suggests one of the supported semantic type for each column.

The semantic type corresponds to the category (names, emails, phone numbers, etc) of the data. If the semantic type that has been applied on a column is not the desired one, you have the possibility to manually change it to one of the predefined types, based on your own experience.

Let's take the example of a dataset containing client data, including the job title of your customers. You can see in the header of the job title column that the data type has only been recognized as String. You are going to change the semantic type of the column so that it more accurately reflects the data.

Note: You can also modify semantic types from the Data model panel of a dataset hierarchical view.


  1. Click the menu icon in the header of the job title column.
  2. From the menu that opens, you can either:
    • Start typing the name of the type that you think would be appropriate in the Find another semantic type field.

      As you type, an auto-completion feature will suggest a list of available types for your data.

    • Select one of the suggestions, based on the matching percentage with your column.
    Note: To change the semantic type in a preparation column, click the menu icon in the header column and click This column is of type to open the semantic type menu.
  3. Click the Job Title type from the suggestions in this case.
    According the statistics, this semantic type corresponds the most to the values contained in the column.
  4. Click Apply changes.


The column type is updated to Job Title, as you can see in the header of the job title column.

Every time that the semantic type of a column is modified, the dataset quality is calculated again.