Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Xianhua Wang

NAME:
- Xianhua Wang

DEGREE:
- Ph.D.

DISCIPLINE:
- Library and Information Science

SCHOOL:
- University of Maryland (USA) (1990)

ADVISORS:
- Dagobert Soergel

COMMITTEE MEMBERS:
- Gary Marchionini
- Elizabeth Aversa
- Marilyn Domas White
- Ben Abraham Shneiderman

MPACT Status: Fully Complete

Title: Knowledge-based selection of databases: An algorithm and its evaluation

Abstract: This dissertation addresses database selection, an issue of increasing importance as computerized files proliferate. Database guides and selection programs are limited in considering full range of capabilities of databases. This project designed and tested the effectiveness of a prototype system to select databases for queries in business. The system provides extensive descriptions of 26 business-related databases, using the entity-relationship approach as the conceptual schema. Data were derived from vendor descriptions and business thesauri and dictionaries. The knowledge basis consists of a conceptual schema, a thesaurus, and the database descriptions. The system is implemented in Prolog. No interface was established for end-user searching. An emphasis in evaluation is the effectiveness of the conceptual schema for representing database information.

To test the system effectiveness, the system's response was compared with databases chosen by professional searchers. Four searchers and one judge participated in the experiment. All were experienced business librarians using a range of online databases. Each searcher chose databases for twelve actual reference queries and stated their reasoning. Using search algorithms optimizing recall, the system generated responses for all 24 questions. For each set of questions, databases selected by the system and each searcher were considered relevant. The judge decided the relevance of databases not unanimously chosen.

The average number of databases chosen by the system and by the searchers was similar, but the median number was less for the system. Overall recall and precision was similar for both. While the agreement between the searchers was about 44%, the agreement between the system and searchers ranged from 34 to 49%. In the prototype, the system performs almost as well as a searcher.

Formative evaluation of the search results, including the searcher's reasoning, provided the basis for determining factors contributing to the success or failure of the prototype. Reasons for failure included, for example, inability to represent specific data, insufficient database descriptions, problems with question interpretation, breadth of searching (algorithm problems). Suggestions for improvement of all components of system are included in the results. Appendices include the formative analysis for each query, the Prolog program, and individual database descriptions, which reflect the conceptual schema.

MPACT Scores for Xianhua Wang

A = 0
C = 0
A+C = 0
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2008-01-31 06:18:02

Advisors and Advisees Graph