Shared utility for preprocessing training data for MT #260
Description
Machine and silnlp have two different implementations of preprocessing/preparing the data for training. We should consider creating a utility function in machine.py that could be used in both machine.py/Serval and silnlp to ensure consistency.