How to Scale the Production of a Large Data Set

This is sort of much more work that's much further out. But I would be interested in scaling the production of such a data set, being able to rapidly build up a large test set for either for an expert system or for a legal reasoning model on a large set of statutes. This could be done automatically, semi-automatically. For this, here, we had to control every aspect of it,. We couldn't simply go out and pick the number of cases from case.law and then, heuristically match them to legal statutes  and then call it a data set. So you have this clean data set now, and then you can make better methods with noisy data