Instead of using an observer of PlatformSpeechSynthesizer when doing speech synthesis in the UI process, it would be more elegant to pass completion handlers to the various speech synthesis calls.
<rdar://problem/48931062>