Text this: Source Modeling Techniques for Quality Enhancement in Statistical Parametric Speech Synthesis