Click here to visit the SIG website!
Aims. The SYNSIG Special Interest Group was formed to promote activities related to furthering the science of speech synthesis. The SIG brings together colleagues with a special interest in Speech Synthesis. It promotes interest in speech synthesis including text-to-speech conversion, concept-to-speech conversion and related disciplines, provides a means of exchanging news on recent developments in research and applications through a web site and the organisation of meetings, workshops and other activities (for instance evaluation), and makes available relevant resources for education.
Motivation. At an international conference on speech processing, a speech scientist once held up a tube of toothpaste (whose brand was "Signal") and, squeezing it in front of the audience, coined the phrase "This is speech synthesis; speech recognition is the art of pushing the toothpaste back into the tube." One could turn this very simplistic view the other way round: users are generally much more tolerant of speech recognition errors than they are willing to listen to unnatural speech. There is magic in a speech recognizer that transcribes continuous radio speech into text with a word accuracy as low as 50%; in contrast, even a perfectly intelligible speech synthesizer is only moderately tolerated by users if it delivers nothing more than "robot voices". Delivering both intelligibility and naturalness has been the holy grail of speech synthesis research for the past 30 years. More recently, expressivity has been added as a major objective of speech synthesis. Add to this the engineering costs (computational cost, memory cost, design cost for making another synthetic voice or another language) which have to be taken into account, and you'll start to have an idea of the challenges underlying text-to-speech synthesis.