2 multi class labels - should I combine them into one?

For my input data I have 2 multi class labels. One is for country - the text can only come from from one country out of many. The other is sentiment, either positive or negative but not both or neither.
I am using Bert and I am wondering if the 2 labels should by input as a 2x1 vector, or if they should be combined into one scalar. For example for UK positive and UK negative where the id for UK is 3, then the input labels would be 3.1 for UK positive and 3.0 for UK negative. If a vector it could be (3, 0) and (3, 1)