Many context-aware computing applications form inferences and execute
corresponding actions based on context that is uniquely associated with a
user. We refer to such applications as customized context-aware
applications, or deep context-aware applications, and recognize that
their design poses a very challenging burden to application designers due
to the degree of customization that is required. To tackle this problem,
we have developed a programming model and framework for context-aware
applications with the goal of shielding application developers from the
complexity of customization. The framework applies machine learning in
novel ways to infer application triggering conditions. In an effort to
evaluate the Context Tailor model we have created an example deep
context-aware application and designed a method for evaluating its
effectiveness both quantitatively and subjectively. In this talk I will
describe the Context Tailor project and the ongoing usability study of
the Smart Invoker application.
DVSC.