We conclude with a summary of what we have learnt from the work on Emacspeak. Firstly, the design of Emacspeak as a speech interface as opposed to a system that reads the screen is radically different from what has been attempted in the past. The current implementation has achieved a remarkable level of success in providing fluent speech access to day-to-day computing tasks.
The convoluted interfaces provided by screen-readers proved moderately effective in the case of visually impaired users -there was no other choice and consequently, users had the motivation to learn and use these interfaces. However, general users who wish to use speech as an extra modality to enhance their interaction with the computer are unlikely to put up with such interfaces. The direct access provided by the speech-enabling approach is likely to produce more acceptable output and make deploying speech interfaces easier.
Finally, our implementation of Emacspeak has provided the first true speech access interface to UNIX workstations. To date, the only available solution for visually impaired users has been to access these using PC's running screen-readers as a talking terminal. Our work provides a viable alternative to accessing the power of UNIX and the wealth of communication and development tools that are commonplace in this environment.