Computer-assisted conversation for nonvocal people using prestored texts

The challenges faced by people who are severely physically impaired and nonspeaking are demonstrated by the fact that, even with current technology, the word rates that they can achieve using current communication systems tend to range from two to ten words per min. We have been investigating ways in which prestored reusable texts might assist in improving the performance of such systems. Beginning with a study of the acceptability and effects of including prestored texts in a conversation, this investigation examined a series of methods of increasing complexity that had the system itself perform some of the cognitive and manipulative tasks required to locate suitable texts for a user to include in conversation. These methods have included logging conversational paths for potential reuse in future interactions, using a hypertext structure for storing and navigating through conversational texts, employing fuzzy information retrieval, and building a system based on scripts, plans, and goals. Seeing the possible conversation modes as occurring at varying levels of novelty, from predictable routines through reusable texts to uniquely created utterances, the most likely area for significant improvement then is in the storage and retrieval of large amounts of reusable text. Research needs to discover ways of making the retrieval process as effortless and efficient as possible to free the user to concentrate on the interaction itself