NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED552030
Record Type: Non-Journal
Publication Date: 2012
Pages: 213
Abstractor: As Provided
Reference Count: N/A
ISBN: 978-1-2678-7092-6
ISSN: N/A
Personalizing Information Retrieval Using Interaction Behaviors in Search Sessions in Different Types of Tasks
Liu, Chang
ProQuest LLC, Ph.D. Dissertation, Rutgers, State University of New Jersey, New Brunswick
When using information retrieval (IR) systems, users often pose short and ambiguous query terms. It is critical for IR systems to obtain more accurate representation of users' information need, their document preferences, and the context they are working in, and then incorporate them into the design of the systems to tailor retrieval to individual users. The proposed study is to personalize IR systems by tailoring search result content to individual users through the inference of useful documents during their information seeking episode, in different types of tasks. Specifically, this dissertation has two research goals: (1) generate predictive models of document usefulness based on multiple user behaviors as in different types of tasks; (2) generate predictive models of task type through observing users' search behaviors. To address these research goals, this study analyzed data collected in a controlled lab experiment. Thirty-two students were invited to participate in the study, each worked on four search tasks, and these tasks were designed to be different types. During search sessions, all users' interactions were recorded by multiple loggers. Predictive models of document usefulness and task type were generated using various statistical analysis methods. Our results demonstrate that multiple behavioral measures on both content pages and search result pages can be indicators of document usefulness. More importantly, task type affected the relationship between the behavioral measures and document usefulness, and it may therefore be necessary to build task-specific predictive models of document usefulness, which can achieve better prediction accuracy than a non-task specific predictive model. In addition, behavioral measures on within-session level and whole-session levels could be able to generate predictive models of task type. The results improve our understanding on how to infer users' search context information and document usefulness from user behaviors, and then to use this knowledge to improve the information searcher's experience; that is, to make their information search more effective and pleasurable. The research findings have theoretical and practical implications for using behavioral measures and taking account of contextual factors in the development of personalized IR systems. Future studies are suggested for making use of these findings as well as research on related issues. [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: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site: http://www.proquest.com/en-US/products/dissertations/individuals.shtml
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A