Fehime Nihan Cicekli
Electrical Engineering & Computer Science
- Multimedia information retrieval
- Recommendation systems
- Semantic web
Semantic user-profiling: The aim of this research is to provide a methodology for constructing a complete user model using hypernetwork and enable dynamic user profile tailoring based on context. The secondary aim is to illustrate how the proposed methodology can be used in personalized search and profile aggregation. The methodology is based on modelling the individual as hypernetwork with a multi-level design. In hypernetwork, the topology creates measurable distances between vertices. Q-Analysis technique is used to calculate the connectivity of vertices with each other and eccentricity measures how much new information the sub-component adds to the structure. We used eccentricity as a threshold to determine the clustering level. The hypernetwork is clustered using Q-Analysis and eccentricity from bottom to top resulting in a coarse structure which corresponds to a summary of user’s profile. This summary structure is used as a guide to tailor user model dynamically based on the provided context.
The methodology is evaluated against one month long Yandex search logs which contains over 167 million records and improved Yandex’s non-personalized ranking. Besides, a small dataset which contains users’ several social accounts is collected to analyze aggregation behavior. The analysis showed that the coverage of aggregated user profile depends heavily on the nature of used social accounts. It is suggested that using proposed methodology enables applying personalization by tailoring the user model based on query and simply using basic heuristics. Alternative approaches are employing statistical or machine learning methodologies which require training the model with existing data.
- Database management systems
- Data structures
- Design and analysis of algorithms
- Information Retrieval
- Guder, M., N.K. Cicekli, Multi-modal Video Event Recognition based on Association Rules and Decision Fusion, Accepted for publication in Multimedia Systems, 2017, DOI: 10.1007/s00530-017-0535-z
- Sinaci, A.A., G.B.L. Erturkmen, S. Gönül, M. Yuksel, A. Pacaci, N. Cicekli, Postmarketing Safety Study Tool: A Web Based, Dynamic, and Interoperable System for Postmarketing Drug Surveillance Studies, BioMed Research International, May 18, 2015.
- Yilmazturk, M.C, I. Ulusoy, N.K. Cicekli, Online annotation of faces in personal videos by sequential learning, Multimedia Tools and Applications: Volume 63, Issue 3 (2013), Page 591-613.
- Kara, S., O. Alan, O. Sabuncu, S. Akpınar, N.K. Cicekli, F.N. Alpaslan, An Ontology-Based Retrieval System Using Semantic Indexing, Information Systems, Volume 37, 2012, pp. 294-305, DOI: 10.1016/j.is.2011.09.004
- Kuzu, M. and N.K. Cicekli, Dynamic Planning Approach to Automated Web Service Composition, in Applied Intelligence, Volume 36, pp. 1-28, 2012, DOI: 10.1007/s10489-010-0238-z.