Interaction Through Non-interaction: Context Awareness and Distributed Applications

Spoken interaction with machines is not yet commonplace, despite extensive research and some commercial successes. Systems that allow unlimited conversation with arbitrary speakers are still to be seen. In addition to computational difficulties, environmental noise, social issues, multiple languages, and the general ambiguity of human dialogue, one fundamental aspect behind the problem is the generic difficulty of determining the space of discourse. To limit this semantic search space, we hope to get a little help from the contextual info available in ubiquitous computing environments. We discuss some developments in the areas of user profiling, context awareness, sensor networks, social clues, and semantic networks, together with some expectations of how the gained additional information can enhance spoken interaction. We then look at some applications made possible by these developments and suggest topics for further research.