Dissertation Information for Xiaoya Tang NAME: - Xiaoya Tang
DEGREE:
- Ph.D.
DISCIPLINE:
- Library and Information Science
SCHOOL:
- University of Illinois, Urbana-Champaign (USA) (2007)
ADVISORS: - Linda C. Smith
COMMITTEE MEMBERS: - Michael B. Twidale - P. Bryan Heidorn - Chen Xiang Zhai
MPACT Status: Fully Complete
Title: Extraction and use of structured information in full-text retrieval: A case study
Abstract: Most current full-text retrieval systems use statistical methods based on term occurrence in automatic document analysis and indexing. With these approaches keywords and statistical phrases have been the most widely used index terms, and the search process involves matching occurrences of search terms with index terms. While this has proven effective on small surrogate collections, frequency-based terms become too general or ambiguous in specialized full-text collections. As a result, retrieval using such approaches on specialized full-text collections often suffers from poor precision.
This study uses information extraction techniques to automatically generate structured semantic information from documents by applying machine learning algorithms, knowledge bases, and partial "understanding" techniques to documents. The extracted information is used to represent document content in a way that is more specific, accurate, and meaningful, and is also more usable by real users. An interface is provided to allow users to use such information to form search queries, and a matching mechanism and weighting strategy are provided to match user queries with documents.
An experiment involving real users was conducted to evaluate this approach when applied to the full-text botanical collection Flora of North America (FNA). The experimental results indicate that the structured semantic information improves retrieval performance by allowing the users to accomplish both more information tasks per unit time and more successful searches per unit time. Users also reported more satisfaction with the completeness of the retrieval results, better ease of use, and greater overall satisfaction.
This work contributes to information retrieval research by providing a more specific and accurate document content representation and consequently more effective full-text retrieval and by providing a better understanding of how structured semantic information can impact full-text retrieval performance. This study provides a practical approach for extracting and using structured semantic information in real retrieval systems that can be generalized for application to similar data collections or other special collections and improves the usability of such large full-text databases.
| |
MPACT Scores for Xiaoya Tang A = 0
C = 0
A+C = 0
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2009-06-03 20:17:26
Advisors and Advisees Graph
generating graph, please reload |