RecSys Challenge 2024

Balancing Accuracy and Editorial Values
in News Recommendations

Sponsored by


About

This year's challenge focuses on online news recommendation, addressing both the technical and normative challenges inherent in designing effective and responsible recommender systems for news publishing. The challenge will delve into the unique aspects of news recommendation, including modeling user preferences based on implicit behavior, accounting for the influence of the news agenda on user interests, and managing the rapid decay of news items. Furthermore, our challenge embraces the normative complexities, involving investigating the effects of recommender systems on the news flow and whether they resonate with editorial values.

Register

Challenge Task

The Ekstra Bladet RecSys Challenge aims to predict which article a user will click on from a list of articles that were seen during a specific impression. Utilizing the user's click history, session details (like time and device used), and personal metadata (including gender and age), along with a list of candidate news articles listed in an impression log, the challenge's objective is to rank the candidate articles based on the user's personal preferences. This involves developing models that encapsulate both the users and the articles through their content and the users' interests. The models are to estimate the likelihood of a user clicking on each article by evaluating the compatibility between the article's content and the user's preferences. The articles are ranked based on these likelihood scores, and the precision of these rankings is measured against the actual selections made by users.

Evaluation

To evaluate the models, we use several standard metrics in the recommendation field, including the area under the ROC curve (AUC), mean reciprocal rank (MRR), and normalized discounted cumulative gain (nDCG@K) for K shown recommendations. To address the normative complexities inherent in news recommendations, the test set incorporates samples specifically designed to assess models based on normative properties. This includes evaluating models on Beyond-Accuracy Objectives, such as intra-list diversity, serendipity, novelty, coverage, among others. The final result is the average of these metrics across all impression logs.

  • The primary metric for the challenge is AUC.
  • Prizes

    The top three teams will receive exciting cash prizes: $3,500 for first place, $2,500 for second, and $1,500 for third. Additionally, a special $2,500 prize will be awarded to the best academic team.

    Timeline

    The table below presents the comprehensive timeline and critical deadlines relevant to the challenge. It's essential to note that all listed dates and times are based on the Anywhere on Earth (AoE) timezone, marked at 23:59:59.

    When? What?
    Mar. 8, 2024 Start RecSys Challenge
    Release dataset
    Mar. 25, 2024 Submission System Open
    Apr. 4, 2024 Leaderboard live
    Jun. 21, 2024 End RecSys Challenge
    Jun. 24, 2024 Final Leaderboard & Winners
    EasyChair open for submissions
    Jul. 1, 2024 Code Upload
    Upload code of the final predictions
    Jul. 18, 2024 Paper Submission Due
    Aug. 3, 2024 Paper Acceptance Notifications
    Aug. 29, 2024 Camera-Ready Papers
    Oct. 14, 2024 RecSys Challenge Workshop
    @ACM RecSys 2024

    Organizers

    The challenge is organized by Ekstra Bladet and JP/Politikens Hus A/S ("Ekstra Bladet"), Johannes Kruse1,2, Kasper Lindskow1, Anshuk Uppal2, Michael Riis Andersen2, Jes Frellsen2, Marco Polignano3, Claudio Pomo4 and Abhishek Srivastava5 based on the data provided by Ekstra Bladet.
    1. Ekstra Bladet / JP/Politikens Hus A/S
    2. Technical University of Denmark
    3. University of Bari Aldo Moro, Italy
    4. Politecnico di Bari, Italy
    5. IIM Visakhapatnam, India
    If you have any questions, suggestions, or are experiencing any issues, do not hesitate to reach out:
    • Johannes.Kruse@jppol.dk


    Contributions

    Accepted Papers

    Guidelines

    We invite researchers and practitioners to submit their work for the RecSys Challenge workshop. Note that winning teams must submit papers and sign up for the workshop.

    Starting this year, the RecSys submissions have adopted a new template. We will follow the same new rules that apply to all other types of papers, please follow the Call for Paper Submission Guidelines.

    The topics of interest include, but are not limited to (in alphabetical order):

    • Applications of news recommendation
    • Benchmarking and evaluation of recommender systems
    • Bias in intelligent news systems
    • Clickbait, fake news and misinformation detection
    • Contributions focused on beyond accuracy, such as fairness, diversity, coverage, etc.
    • Cross-domain and multi-modal recommendations
    • Dataset analyses and preprocessing techniques
    • News categorization, summarization and headline generation
    • News content modeling
    • News ranking techniques
    • News trend and lifecycle
    • Novel model architectures for news recommendation
    • Privacy protection in news recommendation
    • Scalability and efficiency of recommendation algorithms
    • User behavior analysis
    • User interest modeling
    Furthermore, we ask that solutions using all features, including those that may yield information not available in a live setup, and report results both with and without these features (as discussed in the thread: link).


    Program

    The workshop will take place on Monday, October 14, 2024, at Politecnico di Bari in Aula Magna on the second floor. The program is organized into three focused sessions, covering topics such as user modeling, beyond-accuracy objectives, application of machine learning techniques and traning strategies. Between sessions, we will have breaks for networking, giving attendees an opportunity to connect and collaborate. Each talk is allocated 15 minutes for the presentation, followed by 5 minutes for audience questions and discussion.

    For a detailed look at the entire conference, including other tracks and keynotes, see the full conference schedule. Below is the complete agenda for our workshop:

    Workshop Presentations

    Workshop Pictures

    Session 1: 09:00 - 10:30
    1 09:00 - 09:15 Opening Remarks
    2 09:15 - 10:00 Keynote Speech: Balancing Accuracy and Editorial Values in News Recommendations
    Kasper Lindskow, Ph.D., Head of AI at JP/Politikens Media Group
    3 10:00 - 10:15 Leveraging User History with Transformers for News Clicking: The DArgk Approach
    Juan Manuel Rodriguez and Antonela Tommasel
    4 10:15 - 10:30 Recommendations for the Recommenders: Reflections on Prioritizing Diversity in the RecSys Challenge
    Lucien Heitz, Sanne Vrijenhoek, and Oana Inel
    Coffee Break: 10:30 - 11:15
    Session 2: 11:15 - 12:45
    5 11:15 - 11:35 Exploiting Contextual Normalizations and Article Endorsement for News Recommendation (🥇)
    Andrea Alari, Lorenzo Campana, Federico Giuseppe Ciliberto, Saverio Maggese, Carlo Sgaravatti, Francesco Zanella, Andrea Pisani, and Maurizio Ferrari Dacrema
    6 11:35 - 11:55 Harnessing Temporal Dynamics and Content: An Ensemble of Gradient Boosting Machines for News Recommendation (🥉)
    Tomomu Iwai, Akihiro Tomita, Tomoyuki Arai, Hiroki Ogawa, and Takuma Saito
    7 11:55 - 12:15 Large Scale Hierarchical User Interest Modeling for Click-through Rate Prediction (🥈)
    Taofeng Xue, Zhimin Lin, Zhijian Zhang, Linsen Guo, Haoru Chen, Mengjiao Bao, and Peng Yan
    8 12:15 - 12:35 Enhancing News Recommendation with Transformers and Ensemble Learning (🥇)
    Kazuki Fujikawa, Naoki Murakami, and Yuki Sugawara
    Lunch Break: 12:45 - 14:30
    Session 3: 14:30 - 16:00
    9 14:50 - 15:10 Enhancing News Recommendation with Real-Time Feedback and Generative Sequence Modeling
    Qi Zhang, Jieming Zhu, Jiansheng Sun, Guohao Cai, Ruining Yu, Bangzheng He, and Liangbi Li
    10 15:10 - 15:30 DIVAN: Deep-Interest Virality-Aware Network to Exploit Temporal Dynamics in News Recommendation
    Antonio Ferrara, Marco Valentini, Paolo Masciullo, Antonio De Candia, Davide Abbattista, Riccardo Fusco, Claudio Pomo, Vito Walter Anelli, Giovanni Maria Biancofiore, Ludovico Boratto, and Fedelucio Narducci
    11 15:30 - 15:50 Leveraging LightGBM Ranker for Efficient Large-Scale News Recommendation Systems
    Tetsuro Sugiura, Yosuke Yamagishi, and Yodai Kishimoto
    12 14:30 - 14:50 An User Interest Modeling with Diverse Behavior Analyses and Embeddings for Building Online News Recommendation Systems
    Jiangwei Luo, Ye Tang, Shien Song, and Haibo Lu
    13 15:50 - 16:00 Winners' Ceremony (🥇 🥈 🥉) & Closing Remarks
    Announcement of the top three teams and the best academic team