Project MULTIVIEW

Januar 2025

Tanise Ceron, Amelie Wuehrl

MULTIVIEW develops methods and models for balanced news recommendation.

Vector Stiftung and Software Campus

Recommendation algorithms have a profound impact on democracies: they can boost or hinder the interaction of citizens with several opinions and viewpoints. In this project we develop methods to reshape news recommendation. Using AI, we automatically identify perspectives in news articles and develop a recommender that provides a wide view of perspectives to users. This allows, for example, news consumers to have a broader understanding of events or current affairs in the world.

Members of the project:

Dr. Hardy - postdoctoral researcher

Agnese Daffara - PhD researcher 

Master theses:

• Seemab Hassan. 2025. Assessing Large Language Models (LLMs) for Frame Prediction: A Study on Generalization Across Diverse Political Issues and Communication Contexts 
• Sourabh Ramesh Dattawad. 2025. Integrating News Frames into Recommender Systems: Effects on Normative Diversity and Quality of News Recommendations
• Xiaochen Jia. In progress. Incorporating Entity-Level Sentiment to Promote Normative Diversity in News Recommendations

Publications:

• Hardy, Sebastian Padó, Amelie Wührl, and Tanise Ceron. 2025. Democratizing News Recommenders: Modeling Multiple Perspectives for News Candidate Generation with VQ-VAE. arXiv preprint arXiv:2508.13978.
• Agnese Daffara, Sourabh Dattawad, Sebastian Padó, and Tanise Ceron. 2025. Generalizability of Media Frames: Corpus creation and analysis across countries. In Proceedings of the 14th Joint Conference on Lexical and Computational Semantics (*SEM 2025), pages 83–99, Suzhou, China. Association for Computational Linguistics.
• Sourabh Dattawad, Agnese Daffara, and Tanise Ceron. 2025. Leveraging Media Frames to Improve Normative Diversity in News Recommendations. arXiv preprint arXiv:2509.02266.