For the source view, BigML quickly peeks into a small portion of the data to get a preview of the kind of data you have uploaded. Whereas datasets analyze and process all the data uploaded to BigML. At this point, when a field is categorical yet the number of unique values is too large, we subsequently consider it an unstructured text field.
If you have a text field that has been automatically identified as categorical after uploading your data to BigML, you will need to correct the type. To fix it, just go to your source and click on the Configure source menu option:
Once in the source details panel, change the type of the field from categorical to text, and click the Update button.
The field type is also updatable from the BigML API, as explained in our documentation.
Notice that if you have categorical fields with more than 1,000 unique values, you will not be able to set them as categorical, as this is the maximum number allowed. However, this limit is fully customizable with a BigML Private Deployment.