Using casual speech phonology in synthetic speech

Alphabetic writing is a mixed blessing for speech science. Most scientists working in speech synthesis and speech recognition assume unconsciously that spoken language is like written language, i.e. it is composed of a string of items (letters/phonemes) which should be realised in all but substandard writing/speech. My research shows that there are very many shortcuts taken by speakers of English on a regular basis in normal (not sloppy or casual) speech. These are not included in speech synthesis packages, but if they were, the output would be closer to the real thing and, I contend, would be considerably easier to understand.