You are in: Submissions > Select institution > University of Warwick > UOA 23 - Computer Science and Informatics > RA1, RA2 and RA5c

University of Warwick

UOA 23 - Computer Science and Informatics

RA1, RA2 and RA5c: Staff and output details and Category C staff circumstances


Anand, S S - Category : A

RA2 - Research outputs:

Number of outputs: 3

Output number: 1 of 3


Generating Semantically Enriched User Profiles for Web Personalization

Output type: Journal article
Journal title: ACM Transactions on Internet Technology
Month/year of publication: October 2007
Pagination: 22 (1-26)
Volume: 7 (4)
ISSN: 1533-5399
URL: Original article on web ?
DOI: 10.1145/1278366.1278371 ?
Co-authors: Kearney, P., Shapcott, M.
Other relevant details: This paper appeared in a special issue on of ACM TOIT on Web Personalization, with a 22% acceptance rate. The ACM TOIT journal is considered to be one of the top journals for the field. The paper presents a novel method for generating recommendations that addresses two important problems in recommendation systems: incorporating item semantics and rating sparsity. Central to the technique is the determining of hidden factors that drive user ratings within an interaction with the recommender system. The results showed a 25% increase in recall by the method proposed in the paper over standard recommendation systems.

Output number: 2 of 3


Contextual Recommendation

Output type: Chapter in book
Editors: B Berendt, A Hotho, D Mladenic, G Semeraro
Book title: From Web to Social Web: Discovering and Deploying User and Content Profiles
Publisher: Springer Berlin / Heidelberg
Year of publication: 2007
Pagination: 142-160
ISBN: 978-3-540-74950-9
Co-authors: Mobasher, B
Other relevant details: This is one of only two papers, out of seven, from the Workshop on Ubiquitous Knowledge Discovery for Users invited to be extended and published within this book. The paper revisits the modelling of users within recommender systems specifically looking at the role of context within recommendation and how such information can be incorporated within the user model. A user model based on research in cognitive science was presented and used for recommendation generation. An increase in recall of 18-26%, over standard recommendation techniques, depending on the amount of data collected in the current user interaction, was reported.

Output number: 3 of 3


DIVA: A variance-based clustering approach for multi-type relational data

Output type: Conference contribution
Conference: Proceedings of the 16th ACM Conference on Information and Knowledge Management (CIKM 2007): session B2 on Classification and Clustering I.
Month/year of publication: 06/11/2007
Number of pages: 147-156
Media of output:
URL: Original article on web ?
Co-authors: Li, T.
Other relevant details: The paper presents a new approach to clustering objects in multi-relational space. The algorithm uses the concept of representative objects to estimate the mediods of the clusters to reduce its computational complexity to O(N). Extensive experiments on real and synthetic data sets show that the algorithm improves on previously proposed clustering techniques in both, scalability as well as accuracy. The paper also proposes an alternative to the need for users to specify the number of clusters expected in the data, using a variance bound instead to estimate the number of clusters. The conference has a 17% acceptance rate.