Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Look Into News Avoidance Through AWRS: An Avoidance-Aware Recommender System

Igor L. R. Azevedo, Toyotaro Suzumura, Yuichiro Yasui

TL;DR

This work tackles news article avoidance in recommender systems by introducing AWRS, an Avoidance-Aware Recommender System that jointly models exposure, avoidance, and relevance. The framework integrates a document encoder, an avoidance-aware relevance predictor, and an avoidance-aware user encoder, augmented by a user engagement embedding derived from a grid of exposure and avoidance signals. Evaluated on three multilingual datasets (English, Norwegian, Japanese), AWRS consistently outperforms strong baselines across AUC, MRR, and nDCG metrics, demonstrating the value of incorporating avoidance information into personalization. The findings highlight that avoidance signals reveal nuanced user preferences and, when combined with time-aware relevance, enhance the accuracy and robustness of news recommendations in dynamic environments.

Abstract

In recent years, journalists have expressed concerns about the increasing trend of news article avoidance, especially within specific domains. This issue has been exacerbated by the rise of recommender systems. Our research indicates that recommender systems should consider avoidance as a fundamental factor. We argue that news articles can be characterized by three principal elements: exposure, relevance, and avoidance, all of which are closely interconnected. To address these challenges, we introduce AWRS, an Avoidance-Aware Recommender System. This framework incorporates avoidance awareness when recommending news, based on the premise that news article avoidance conveys significant information about user preferences. Evaluation results on three news datasets in different languages (English, Norwegian, and Japanese) demonstrate that our method outperforms existing approaches.

A Look Into News Avoidance Through AWRS: An Avoidance-Aware Recommender System

TL;DR

This work tackles news article avoidance in recommender systems by introducing AWRS, an Avoidance-Aware Recommender System that jointly models exposure, avoidance, and relevance. The framework integrates a document encoder, an avoidance-aware relevance predictor, and an avoidance-aware user encoder, augmented by a user engagement embedding derived from a grid of exposure and avoidance signals. Evaluated on three multilingual datasets (English, Norwegian, Japanese), AWRS consistently outperforms strong baselines across AUC, MRR, and nDCG metrics, demonstrating the value of incorporating avoidance information into personalization. The findings highlight that avoidance signals reveal nuanced user preferences and, when combined with time-aware relevance, enhance the accuracy and robustness of news recommendations in dynamic environments.

Abstract

In recent years, journalists have expressed concerns about the increasing trend of news article avoidance, especially within specific domains. This issue has been exacerbated by the rise of recommender systems. Our research indicates that recommender systems should consider avoidance as a fundamental factor. We argue that news articles can be characterized by three principal elements: exposure, relevance, and avoidance, all of which are closely interconnected. To address these challenges, we introduce AWRS, an Avoidance-Aware Recommender System. This framework incorporates avoidance awareness when recommending news, based on the premise that news article avoidance conveys significant information about user preferences. Evaluation results on three news datasets in different languages (English, Norwegian, and Japanese) demonstrate that our method outperforms existing approaches.
Paper Structure (34 sections, 13 equations, 8 figures, 6 tables)

This paper contains 34 sections, 13 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction.
  2. Related Work.
  3. Principal Elements.
  4. Exposure.
  5. Avoidance.
  6. Consequences of Avoiding News Articles.
  7. Visualization of Avoidance.
  8. Relevance.
  9. User Engagement Embedding.
  10. Methodology.
  11. Problem Formulation.
  12. Document Encoder.
  13. Avoidance-aware Relevance Predictor.
  14. Avoidance-aware User Encoder.
  15. Historical and Candidate News Vector Concatenation.
  16. ...and 19 more sections

Figures (8)

  • Figure 1: Avoidance and exposure explanation diagram
  • Figure 2: $Av(n, t) \text{ vs. } EPI(n, t)$ for MIND-smallwu-etal-2020-mind at 2019-11-09 from 00:00 AM to 08:00 AM.
  • Figure 3: The graph of $Av(n, t)$ versus $EPI(n, t)$ is generated for the MIND-smallwu-etal-2020-mind (a), Adressa one-week10.1145/3106426.3109436 (b), and Nikkei one-week (c) datasets, using $D = 5$, which results in 25 distinct regions.
  • Figure 4: Overview of AWRS.
  • Figure 5: (a) Avoidance-aware Relevance Predictor and (b) Avoidance-aware User Encoder Schematics.
  • ...and 3 more figures

Theorems & Definitions (4)

  • Definition 1: Number of Exposures - $n_E(n, t)$
  • Definition 2: Number of Impressions - $n_I(t)$
  • Definition 3: Exposure Per Impression - $EPI(n, t)$
  • Definition 4: Avoidance - $Av(n, t)$