Using Collaborative Filtering to Weave an Information Tapestry

摘要

Tapestry is an experimental system that manages an incoming stream of electronicdocuments, including electronic mail, news wire stories and NetNews articles. In common with some recent mail systems, Tapestry uses filtering to cope with large volumes of incoming documents. Where Tapestry differs from these systems is in its philosophy that humans provide the most reliable evaluation of documents, and so it uses collaborative filtering, which it implements by having users annotate documents, and then allowing filtering using those annotations. Because annotations are not known at the time that documents arrive, Tapestry filters must not only test incoming documents, but must also run repeatedly over the entire database of documents. The paper is a report on the Tapestry design and the status of its implementation. (Copyright (c) 1992 Xerox Corporation. All rights reserved.)