Dissertation Information for Odette Maroun Badran
MPACT Status: Fully Complete
Title: A probabilistic approach to information retrieval
Abstract: The Boolean technique is the most commonly used method ot retrieving references from contemporary bibliographic databases, while this technique sometimes performs well in terms of recall, its perfornance in terms of precision is much less satisfactory. The typical Boolean search frequently retrieves a large set of references, many of which are irrelevant to the user's information need. To overcome this problem three additional retrieval techniques were developed to filter irrelevant references from the output of Boolean searches. The experimental techniques assume a relationship among the documents in the file and uses this relationship to accomplish the filtering. Disjoint InterCommunication Classes and Communication Chains are search techniques based on a similarity measure derived from the Indirect Method, a conditional probability model developed by Dr. William C. Coffman. The Normalized Symmetric Difference is based on a dissimilarity measure suggested by C.J. Van Rijsbergen and represents a special case of the Indirect Method. The effectiveness of several modes of content representation -- co-citation, title key-words, and bibliographic cOupling -- as indicators of document relatedness was also tested. The performance of each technique and content mode was evaluated using precision, recall, and a composite measure of effectiveness. Disjoint lntercomunication Classes and co-citation were found to produce the best results. When applied to tne output of a Boolean search, they improve output by filtering the number of irrelevant references retrieved.
MPACT Scores for Odette Maroun Badran
Advisors and Advisees Graph
Students under Odette Maroun Badran