Article,
Detecting Contexts on the Desktop Using Bayesian Networks
Deskt. Search Work. SIGIR2010, (2010)
Abstract
A good understanding of a user?s (working) contexts pro- vides the basis for improved desktop information management, as well as for personalized desktop and Web search. We propose to combine a variety of evidences found by analyzing desktop information for inferring the user?s working contexts, and more specifically infer file-to-context assignment using a Bayesian Network. Our preliminary experiments focus on identifying a good selection of evidences to use and show that the choice of evidences is coherent with user assessments for desktop files, as well as the contexts inferred by the Bayesian Network.
