Deriving Vocal Interfaces from Logical Descriptions in Multi-device Authoring Environments