| Abstract: |
The integration of query generation and user feedback continues to
challenge information retrieval technologies. Relevance feedback, while useful to
professionals, is frequently inappropriate for lay users, because the initial query
generation is manual and the subsequent feedback solicitation is intrusive or inconsistent
with many lay users' information needs. To provide lay users with an integration of
automatic query generation and nonintrusive feedback solicitation, the concept of an
application-embedded information retrieval system is developed. Such a system allows the
application in which it is embedded to become part of an infrastructure of distributed
information sources. Background samples of the application's usage are collected and used
in retrievals from the sources. Feedback is never solicited explicitly, and is utilized
only when volunteered. Retrieval is adjusted through background sampling, anydata
indexing, and dual space feedback. The approach is illustrated with a system embedded in
Gnu Emacs , the Free Software Foundation's text editor. |