This gives me an idea to make a Rosetta Stone but not for languages but for Models (SVM / Gradient Boosting / Linear Regression...). Probably would have to do it using two datasets or more.