Generate predictions for responses with non-numeric human scores

[by @aoifecahill ] It would be nice if it was possible to automatically generate predictions for data that has non-numeric human scores. Obviously these predictions can't be used in most of the evaluations, but they could be useful in a confusion matrix for example.

I know RSMPredict does predictions, but that's not quite what I was looking for. It would be nice if there was just one tool that trained a model using the training file and then ran RSMPredict on the test file. I don't think anything like that already exists? I don't really care much about the evaluation (though the confusion matrix would be interesting), rather just would like the convenience of being able to do train+predict in one step rather than multiple with hacks (since I don't know of a way to just train a model without having a fake test file for rsmtool?)

This is a nice-to-have issue.