Parallelism in Speech Synthesis