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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Traditional content-based tag recommender systems directly learn theassociation between user-generated content (UGC) and tags based on collectedUGC-tag pairs. However, since a UGC uploader simultaneously creates the UGC andselects the corresponding tags, her personal preference inevitably biases thetag selections, which prevents these recommenders from learning the causalinfluence of UGCs' content features on tags. In this paper, we propose a deepdeconfounded content-based tag recommender system, namely, DecTag, to addressthe above issues. We first establish a causal graph to represent the relationsamong uploader, UGC, and tag, where the uploaders are identified as confoundersthat spuriously correlate UGC and tag selections. Specifically, to eliminatethe confounding bias, causal intervention is conducted on the UGC node in thegraph via backdoor adjustment, where uploaders' influence on tags leakedthrough backdoor paths can be eliminated for causal effect estimation.Observing that adjusting the causal graph with do-calculus requires integratingthe entire uploader space, which is infeasible, we design a novel Monte Carlo(MC)-based estimator with bootstrap, which can achieve asymptotic unbiasednessprovided that uploaders for the collected UGCs are i.i.d. samples from thepopulation. In addition, the MC estimator has the intuition of substituting thebiased uploaders with a hypothetical random uploader from the population in thetraining phase, where deconfounding can be achieved in an interpretable manner.Finally, we establish a YT-8M-Causal dataset based on the widely usedYouTube-8M dataset with causal intervention and propose an evaluation strategyaccordingly to unbiasedly evaluate causal tag recommenders. Extensiveexperiments show that DecTag is more robust to confounding bias thanstate-of-the-art causal recommenders.

Deep Deconfounded Content-based Tag Recommendation for UGC with Causal Intervention