Knowledge integration for medical informatics: An experiment on a cancer information system
AuthorHouston, Andrea Lynn, 1954-
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractThis research investigated the question of whether automatic or system-generated information classification methods can help humans better manage information. A series of four experiments were conducted; they investigated the usability (i.e., usefulness) of two automatic approaches to information classification, the concept space approach and a Kohonen-based SOM approach in the context of information retrieval. The concept space approach was evaluated in three different domains: Electronic Brainstorming (EBS) sessions, the Internet, and medical literature (the CancerLit collection). The Kohonen-based SOM approach was evaluated in the Internet and medical literature (CancerLit) domains only. In each case, the approach under investigation was compared with existing systems in order to demonstrate performance viability. The basic premise that information management, in particular information retrieval, can be successfully supported by system-based information classification techniques and that humans would find such techniques viable and useful was supported by the experiments. The concept space approach was more successful than the Kohonen-based SOM approach. After modifications to the algorithms based on user feedback from the EBS experiments had been made, users found the concept space approach results to be comparable (in the Internet study) or superior (in the CancerLit study) to existing information classification systems. The key future enhancement will be incorporation of better ways to identify document descriptors through syntactic and semantic front-end processing. The Kohonen-based SOM approach was considered difficult to use in all but one specialized case (the dynamic SOM created as part of the CancerLit prototype). This can probably be attributed to the fact that its associative organization does not match with the standard mental models (hierarchical and alphabetic) for information classification.
Degree ProgramGraduate College