Let's say I have a dataset which contains lots of inputfilename and outputfilename. How should I go ahead if I want to make a machine learning model for this, so that I later on can inject a inputfulename and get a suggestion for outputfilename?