Applying local latent semantic indexing for information retrieval visualization
Date
1997
Authors
Miller, Michael Hugh
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Health professionals and consumers are not keeping pace with advancements in knowledge reported in the scientific literature. One of the reasons for this state of affairs is the lack of effective tools to search and present relevant information from MEDLINE and other literature sources. Current techniques are hindered by the ambiguities of natural language and the difficulty of presenting results from a search to the user effectively. This thesis presents a number of core methodologies for indexing and searching text databases such as MEDLINE and discusses the benefits and drawbacks of each method. Four techniques for visualizing search results are also presented and discussed including a novel method proposed by the author called Local Latent Semantic Indexing / Cluster (LLSI/Cluster). Experiments with LLSI/Cluster on three test collections of MEDLINE articles indicate that similar articles tend to cluster together. These findings suggest that LLSI/Cluster has potential as a visualization method for displaying a large set of documents to the user in a graphical manner.