Papers presented on Recommender Systems @ECIR’22 Confernce

Dominik Kowald and Emanuel Lacic, area managers of the Social Computing research area of Know-Center, will present three papers on recommender systems that were created in course of the European H2020 TRUSTS project at the European Conference on Information Retrieval (ECIR’22) in Stavanger, Norway.

In the first paper [1] that will be presented in the BIAS workshop at ECIR’22, Kowald and Lacic analyzed the effect of popularity bias, i.e., popular items are overrepresented in the recommendation lists, on collaborative filtering-based algorithms using four open multimedia datasets. They found that all analyzed algorithms are prone to popularity bias and that users interested in unpopular items get significantly worse recommendations than users interested in popular items.

This is related to the second paper [2] which will be presented during the main conference days of ECIR’22. Focus of the paper is the analysis of popularity bias in an online evaluation study in the news domain. Popularity bias can be mitigated to some extent if we use personalized, content-based recommendations in contrast to unpersonalized recommendations. However, both papers show that we need to be aware of the potential issue of popularity bias also in course of TRUSTS, in which datasets, services and applications are recommended to users. This could be done by e.g. using content-based methods.

Finally, in the third paper [3] that will be presented in the industry day at ECIR’22, Kowald and Lacic present the scalable multi-domain recommendation framework ScaR, which is also used to build the recommendation services in TRUSTS.

[1] Lacic, E., & Kowald, D. (2022). Popularity Bias in Collaborative Filtering-Based Multimedia Recommender Systems. In Proceedings of the BIAS Workshop co-located with the 44th European Conference on Information Retrieval (ECIR’2022). Springer.

[2] Lacic, E., Fadljevic, L., Weissenboeck, F., Lindstaedt, S., & Kowald, D. (2022). What Drives Readership? An Online Study on User Interface Types and Popularity Bias Mitigation in News Article Recommendations. In Proceedings of the 44th European Conference on Information Retrieval (ECIR’2022). Springer.

[3] Lacic, E., & Kowald, D. (2022). Recommendations in a Multi-Domain Setting: Adapting for Customization, Scalability and Real-Time Performance. In Industry-Day Track of European Conference on Information Retrieval (ECIR’2022)