Text this: Speech recognition in adverse conditions: synthetic transformations and real environmental noise