Parameter Tuning of Reranking-based Diversification Algorithms using Total Curvature Analysis

Puthiya Parambath, S. A. , Liu, S., Anagnostopoulos, C. , Murray-Smith, R. and Ounis, I. (2022) Parameter Tuning of Reranking-based Diversification Algorithms using Total Curvature Analysis. In: 8th ACM SIGIR International Conference on the Theory of Information Retrieval (ICTIR 2022), Madrid, Spain, 11-12 July 2022, ISBN 9781450394123 (doi: 10.1145/3539813.3545135)

[img] Text
273316.pdf - Accepted Version
Available under License Creative Commons Attribution.



In this paper, we analyze re-ranking based recommendation diversification algorithms and observe that, commonly, such algorithms can be unified under the scheme of maximizing submodular or modular objective functions from the class of parameterized concave over modular functions. We showcase that such diversification objective functions can be expressed in a generic functional form consisting of the relevance and diversity terms. We then theoretically analyze and show that the total curvature of submodular functions provides insights about the relevance-diversity trade off. This is expected to support data analysts to seek balanced hyperparameter values and, thus, serve as a 'vehicle of validation' by checking the total curvature of submodular objective functions. Our experimental evaluation and performance assessment over benchmark datasets are aligned with our theoretical analysis. We also discuss the importance of balanced trade-off between relevance and diversity in specific application settings like news recommendations to trade-off algorithmic bias and short term user engagement.

Item Type:Conference Proceedings
Glasgow Author(s) Enlighten ID:Murray-Smith, Professor Roderick and Ounis, Professor Iadh and Anagnostopoulos, Dr Christos and Liu, Siwei and Puthiya Parambath, Dr Sham
Authors: Puthiya Parambath, S. A., Liu, S., Anagnostopoulos, C., Murray-Smith, R., and Ounis, I.
College/School:College of Science and Engineering
College of Science and Engineering > School of Computing Science
Copyright Holders:© 2022 Copyright held by the owner/author(s)
First Published:First published in ICTIR '22: Proceedings of the 2022 ACM SIGIR International Conference on Theory of Information Retrieval
Publisher Policy:Reproduced with the permission of the publisher
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record

Project CodeAward NoProject NamePrincipal InvestigatorFunder's NameFunder RefLead Dept
300982Exploiting Closed-Loop Aspects in Computationally and Data Intensive AnalyticsRoderick Murray-SmithEngineering and Physical Sciences Research Council (EPSRC)EP/R018634/1Computing Science