NotesFAQContact Us
Search Tips
Back to results
ERIC Number: ED555381
Record Type: Non-Journal
Publication Date: 2013
Pages: 184
Abstractor: As Provided
ISBN: 978-1-3034-2862-3
Identifying Experts and Authoritative Documents in Social Bookmarking Systems
Grady, Jonathan P.
ProQuest LLC, Ph.D. Dissertation, University of Pittsburgh
Social bookmarking systems allow people to create pointers to Web resources in a shared, Web-based environment. These services allow users to add free-text labels, or "tags", to their bookmarks as a way to organize resources for later recall. Ease-of-use, low cognitive barriers, and a lack of controlled vocabulary have allowed social bookmaking systems to grow exponentially over time. However, these same characteristics also raise concerns. Tags lack the formality of traditional classificatory metadata and suffer from the same vocabulary problems as full-text search engines. It is unclear how many valuable resources are untagged or tagged with noisy, irrelevant tags. With few restrictions to entry, annotation spamming adds noise to public social bookmarking systems. Furthermore, many algorithms for discovering semantic relations among tags do not scale to the Web. Recognizing these problems, we develop a novel graph-based Expert and Authoritative Resource Location (EARL) algorithm to find the most authoritative documents and expert users on a given topic in a social bookmarking system. In EARL's first phase, we reduce noise in a Delicious dataset by isolating a smaller sub-network of "candidate experts", users whose tagging behavior shows potential domain "and" classification expertise. In the second phase, a HITS-based graph analysis is performed on the candidate experts' data to rank the top experts and authoritative documents by topic. To identify topics of interest in Delicious, we develop a distributed method to find subsets of frequently co-occurring tags shared by many candidate experts. We evaluated EARL's ability to locate authoritative resources and domain experts in Delicious by conducting two independent experiments. The first experiment relies on human judges' "n"-point scale ratings of resources suggested by three graph-based algorithms and Google. The second experiment evaluated the proposed approach's ability to identify classification expertise through human judges' "n"-point scale ratings of classification terms versus expert-generated data. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page:]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site:
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A