For multiprocessing, try to use components from the 'joblib' library, along with custom-implemented worker functions. As we are working with large amounts of data here, try to implement multiprocessing wherever possible for speedup. Additionally, we are open to further suggestions on speeding up performance.
Use double quotes on the outside, single quotes on the inside. Seems picky, but regex becomes way harder when quote types are mixed.