A preprocessing (or text normalization) module is necessary as a front end, since TTS [text-to-speech] systems should in principle be able to read any text, including numbers, abbreviations, acronyms, and idiomatics in any format.