iDEA: Drexel E-repository and Archives >
Drexel Theses and Dissertations >
Drexel Theses and Dissertations >
Recreating popular user-generated tags effectively and efficiently by utilizing crowdsourcing
Please use this identifier to cite or link to this item:
|Title: ||Recreating popular user-generated tags effectively and efficiently by utilizing crowdsourcing|
|Authors: ||Elnatour, Deima|
|Keywords: ||Information science--User education;Social sciences--Data processing;User-generated content|
|Issue Date: ||11-Jul-2011|
|Abstract: ||It is well-known today that not all user-generated tags provide additional information that can be used to further improve web search beyond traditional methods.
A number of studies established that popular tags are most useful for web search. Popular tags are common tags provided independently by a large number of users to describe web sources of interest. However, the same studies concluded that incorporating these tags will not make a measurable impact on search engine performance given the size of the web and the scarcity and distribution of popular user-generated tags across the extended web.
This dissertation is focused on finding a way to create social bookmarking tags efficiently and effectively by utilizing crowdsourcing systems. Crowdsourcing is a
platform in which tasks can be posted that would then be completed at a stated price or reward. This aims to attract users to complete relatively simple tasks that are easier for humans and harder for machines. Crowdsourcing has been widely used by companies and researchers to source micro-tasks requiring human intelligence such as identifying objects in images, finding or verifying relevant information, or natural language processing.
The purpose of the study is to determine whether popular internet bookmarking tags can be recreated through crowdsourcing. Amazon Mechanical Turk, the work
marketplace, was used as a means to conduct an experiment regarding the reproduction of popular tags for a variety of websites using Delicious, a service for storing and sharing bookmarked pages on the internet. The key research questions for the study were examined as a number of factors regarding tag creation including the effectiveness of crowdsourcing in reproducing popular tags, categorizing which tags can be recreated most effectively, and the relationship of worker characteristics and demographics to the effectiveness of producing popular tags.
The results of the study suggest that popular internet bookmarking tags can be recreated effectively through crowdsourcing. Moreover, tag creation effectiveness was significantly higher for tag type “Factual and Subjective” (F & S) than for tag type “Factual” (F). Additionally, other variables were tested to assess their relationship with tag creation effectiveness. Interest in site, familiarity with site, tag creation experience and tag usage experience were significantly related to tag creation effectiveness for some of the sites, although the direction and significance of these relationships was not consistent across all sites included in this study.
This study provides a promising new direction for cheap, fast and effective creation of user-generated tags that would be useful in indexing more of the extended
web and consequently help improve web search. Furthermore, it informs future experimental and micro-task design for creating high quality tags reliably using
|Appears in Collections:||Drexel Theses and Dissertations|
Items in iDEA are protected by copyright, with all rights reserved, unless otherwise indicated.