Data-Driven Techniques in Speech Synthesis