Login

Publications  •  Project Statistics

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

Dissertation Information for Xia Lin

NAME:
- Xia Lin

DEGREE:
- Ph.D.

DISCIPLINE:
- Library and Information Science

SCHOOL:
- University of Maryland (USA) (1993)

ADVISORS:
- Gary Marchionini
- Dagobert Soergel

COMMITTEE MEMBERS:
- Gary Wayne Strong
- Claude E. Walston
- James Allen Reggia

MPACT Status: Fully Complete

Title: Self-organizing semantic maps as graphical interfaces for information retrieval

Abstract: This dissertation investigates the potential of using semantic map displays as browsing aids for information retrieval. A map display is proposed as a graphical representation of a document space. A neural network's self-organizing learning algorithm, Kohonen's feature map, is used to construct the map display. The map displays generated by the algorithm are found to show reasonable pictures of the underlying data: their major concepts, their semantic relationships, and their clusters, groups, and distributions. Such map displays can be used as a graphical interface for a retrieval system, in which users can pose broad queries and then browse for relevant documents in the high recall/low precision results.

As an evaluation of the proposed map display, an experiment was conducted to examine how human subjects generate map displays. Eight subjects, each spending two hours, generated eight different map displays by placing 133 document titles on a large two-dimensional grid. The subjects were found to generate two types of maps: category-based maps and association-based maps. The category-based maps tend to be arranged in columns where categories are represented by more or less distinct groups. The association-based maps tend to maintain clear associations among clusters and groups, but boundaries for the clusters and groups are not clear.

Comparing the map displays generated by the algorithm and by the human subjects, similarities are found in the semantic structures of the map displays, the processes of generating the displays, and the functionalities of the displays. When 68 subjects conducted some pseudo-retrieval tasks on the map displays, no significant difference is found between their performance on the human-generated map displays and on the machine-generated map displays, but a significant difference is found between their performance on these map displays and on a random display. These results suggest that the map displays can provide quick access to documents by virtue of their graphical representation. Results of this dissertation strongly imply that information can be and should be made visible to people who search for it.

MPACT Scores for Xia Lin

A = 1
C = 12
A+C = 13
T = 1
G = 1
W = 1
TD = 1
TA = 0
calculated 2013-12-21 17:01:27

Advisors and Advisees Graph