TABLE OF CONTENTS Because it materializes a human communication action, the finest details and subtleties of speech will never be completely unveiled. A simple tape-recorder is sufficient to capture a speech signal, but its study calls for both physical and linguistic concepts.
Current text-to-speech synthesis systems are able to generate speech with an acceptable quality, if the message is neither too long nor too poor in terms of information content. However, it is worthwhile noting that synthetic speech quality has been constantly improving over the years. The technological developments responsible for this progress have mainly concerned acoustic signal processing and computational power speed-ups, for processing ever-larger quantities of linguistic data.
However, will we ever be able to have a machine read Verlaine's poems convincingly? For sure, the current research fields will have to be complemented by more complex formulations and models of our cognitive system. Meanwhile, it is already possible today, with our current level of knowledge, to have speech synthesis systems produce rather surprising utterances.
