| Summary: | With the growth of the internet and other media of communication, locating information
on the topic of interest is less a problem of finding related documents than
determining which particular documents are valuable. Often, the desired information
is obscured within a long list of resources. Users become inundated with so much
information that the task of sifting through it takes the majority of time on a given
information task. Users look at multiple documents at once to find answers to their
questions, and switch between documents to get the âÂÂcompleteâ picture. New systems
are needed that help users cull through related documents to gain the information
they need.
As a part of the Document Triage Project, we have been looking at ways to help
users in sifting through information. The Document Triage Project is developing tools
to recognize, represent, communicate, and visualize user interest across applications.
The topic of this thesis is recognizing user interest and providing an infrastructure
to represent that interest so that it can be shared across the software applications
involved in triage. Based on this inferred interest, applications can help users in their
triage task by providing visualizations or other functionality. The applications could
involve one or many reading interfaces (e.g., a browser, or an editor), an information
organizing system (e.g., Visual Knowledge Builder) and search interfaces (the
application providing the document collection; e.g., a search engine).
To recognize user interest, data is gathered from the userâÂÂs reading, navigational
and interpretive activities. Algorithms based on statistical models and qualitative analyses of user behavior in triage are used to infer interest. A light-weight infrastructure
called Interest Profile Manager has been developed for the representation
of interest values and the corresponding metadata. Interest Profile Manager also provides
text processing capability, interest analysis functionality, sharing of data across
applications and event propagation.
|
|---|
