SectionPage Future directions logo

Future directions for the Dasher project

The Dasher project is being taken forward by the Inference group in the Cavendish laboratory; other researchers who share our aim of seeing Dasher widely used in the real world are also encouraged to work on Dasher.

The initial focus is the development of Dasher for disabled users; additional goals are the development of the palmtop version; the development of Japanese versions; and the creation of specialized versions for particular computers. In order to take all these goals forward, we have released the Dasher source code as an Open Source project. Our Current core activities are listed on the development page.


Dasher on xybernaut

Some longer-range ideas include:

Controlling Dasher by Breath
Dominic Ford
Controlling Dasher by Buttons
Dominic Ford
Controlling Dasher by Tilt sensors
Linking Dasher to the Owl, a very low-cost eyetracker
`Peano' Fully Two-dimensional Dasher
capable of using both dimensions as information sources.
David MacKay
dash
a mouse-driven command-line interface; a one-handed, keyboardless alternative to tcsh, csh, bash, etc.
Unfortunately the name dash has recently been adopted for another shell in Debian, The Debian Almquist Shell.

Davey Jose, Hanna Wallach
Xybernaut
The Xybernaut company have kindly provided a wearable computer for us to test Dasher on.
Improving eyetracker
Phil Cowans and Tim Hospedales
Linux eyetracker
Tim Hospedales
Self-tuning eyetracker
Phil Cowans
Improved language models
Phil Cowans


Dreams for the future

Hybrid voice-dasher system
Speak into an imperfect speech-recognizer, and watch as its inferences are displayed as predictions; wherever it is not sure what you said, use dasher to steer into the correct sentence. Much easier than having to correct errors by saying further speech-commands!
Hybrid automatic translation-dasher system
Assume we have a poor translation system that translates badly from French to English. An expert has to zip through the translation and clean up errors. This cleaning-up could be done within Dasher, using the output of the translator to define a language model.