The following HTML text is provided to enhance online
readability. Many aspects of typography translate only awkwardly to HTML.
Please use the page image
as the authoritative form to ensure accuracy.
consoles no longer exist. Also, speech synthesis has improved a great deal. The best systemsof which the current Bell Labs system is surely an exampleare entirely intelligible, not only to their creators but also to the general population, and sometimes they even sound rather natural. Here I will give a personal view of where this technology stands today and where it seems to be headed. This assessment distills contributions from participants in the colloquium presentations and discussion, who deserve the credit for any useful insights. Omissions, mistakes, and false predictions are of course my own.
The ongoing microelectronics revolution has created a striking opportunity for speech synthesis technology. The computer whose console was mentioned earlier could not do real-time synthesis, even though it filled most of a room and cost hundreds of thousands of dollars. Almost every personal computer now made is big and powerful enough to run high-quality speech synthesis in real time, and an increasing number of such machines now come with built-in audio output. Small battery-powered devices can offer the same facility, and multiple channels can be added cheaply to telecommunications equipment.
Why then is the market still so small? Partly, of course, because the software infrastructure has not yet caught up with the hardware. Just as widespread use of graphical user interfaces in applications software had to wait for the proliferation of machines with appropriate system-level support, so widespread use of speech synthesis by applications will depend on common availability of platforms offering synthesis as a standard feature. However, we have to recognize that there are also remaining problems of quality. Today's synthetic speech is good enough to support a wide range of applications, but it is still not enough like natural human speech for the truly universal usage that it ought to have.
If there are also real prospects for significant improvement in synthesis quality, we should consider a redoubled research effort, especially in the United States, where the current level of research in synthesis is low compared to Europe and Japan. Although there is excellent synthesis research in several United States industrial labs, there has been essentially no government-supported synthesis research in the United Sates for some time. This is in sharp contrast to the situation in speech recognition, where the ARPA's Human Language Technology Program has made great strides, and also in contrast to the situation in Europe and Japan. In Europe there have been several national and European Community-level programs with a focus on synthesis, and in Japan the ATR Interpreting Telephony Laboratory has made significant investments in synthesis as well as recognition