The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9548

Title:Modeling Changing Interests in the Domain of Electronic News
Date: 1995
Abstract:This thesis covers the development and testing of a system intended to reduce information overload in the domain of electronic news articles. The system is designed to aid the user by learning their interests and adapting its filtering pattern to them. By building on the experience of previous systems an adaptive document clustering system was developed which, using high frequency words, extracted the salient features of different documents and used this to predict how relevant the documents were to the user. Ideas from Natural Language Processing research were considered but were not used because they were both very difficult and required a lot of time and domain knowledge. The results of the test were promising though highlighted a number of problems. A discussion of the future improvements to the system is included.

[Search These Pages] [DAI Home Page] [Comment]