Alan Mick

Abstract

Information retrieval systems for documents normally rely on the use of keywords that describe the text in some fashion or another, or are contained in the text itself, for indexing and searching. These keywords may be associated with standard boolean operators, where presence or absence in the text or text description is used as the truth value, or other oper ators indicating their proximity to one another in the text. Another emerging approach is the use of content or knowledge based indexing and retrieval. In this approach the text is not represented or treated as a collection keywords, rather its meaning or semantic content is abstracted and the meaning is used to search for the text desired. This approach may have several advantages over the standard keyword approach. Both precision and recall of the search may be improved, increasing the likelihood that relevant texts will be found while decreasing the probability of finding irrelevant ones. The knowl edge based approach may also allow more sophisticated query techniques, for instance queries based on the purpose for which the text will be used. This thesis will explore the possibility and usefulness of applying case based reasoning to the problem of text search and retrieval. An easy-to-use expert system for information retrieval that utilizes case-based reasoning to improve, over time, its capability to find those items that are relevant and useful, and only those items that are relevant and useful will be implemented. It will support formulation of a search in an intuitive manner that avoids complicated command syntax and occult operators. It will present retrieved docu ments to the user in a logical, useful way and will allow the user to easily refine his search criteria based on a selection of documents from his original results that he has judged to be good examples of what he is searching for.

Library of Congress Subject Headings

Expert systems (Computer science)--Design; Text processing (Computer science); Information storage and retrieval systems--Design

Publication Date

1994

Document Type

Thesis

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Kazemian, Feredoun

Advisor/Committee Member

Wolf, Walter

Advisor/Committee Member

Anderson, Peter

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: QA76.9.E95 M53 1994

Campus

RIT – Main Campus

Download

Share

COinS