Statistics for Non-morphematic word-formation processes: a multi-level approach to acronyms, blends, clippings and onomatopoeia