English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The general aim of the recommender system is to provide personalizedsuggestions to users, which is opposed to suggesting popular items. However,the normal training paradigm, i.e., fitting a recommender model to recover theuser behavior data with pointwise or pairwise loss, makes the model biasedtowards popular items. This results in the terrible Matthew effect, makingpopular items be more frequently recommended and become even more popular.Existing work addresses this issue with Inverse Propensity Weighting (IPW),which decreases the impact of popular items on the training and increases theimpact of long-tail items. Although theoretically sound, IPW methods are highlysensitive to the weighting strategy, which is notoriously difficult to tune. Inthis work, we explore the popularity bias issue from a novel and fundamentalperspective -- cause-effect. We identify that popularity bias lies in thedirect effect from the item node to the ranking score, such that an item'sintrinsic property is the cause of mistakenly assigning it a higher rankingscore. To eliminate popularity bias, it is essential to answer thecounterfactual question that what the ranking score would be if the model onlyuses item property. To this end, we formulate a causal graph to describe theimportant cause-effect relations in the recommendation process. Duringtraining, we perform multi-task learning to achieve the contribution of eachcause; during testing, we perform counterfactual inference to remove the effectof item popularity. Remarkably, our solution amends the learning process ofrecommendation which is agnostic to a wide range of models -- it can be easilyimplemented in existing methods. We demonstrate it on Matrix Factorization (MF)and LightGCN [20]. Experiments on five real-world datasets demonstrate theeffectiveness of our method.

Model-Agnostic Counterfactual Reasoning for Eliminating Popularity Bias in Recommender System