Abstract
The integration of rich, multimodal signals—spanning visual, textual, and acoustic information—represents a significant evolution for recommender systems, promising more nuanced and personalized user experiences. However, the efficacy and trustworthiness of these advanced models hinge critically on a foundational, yet frequently overlooked, element: the integrity of the input data. Practical deployments are often plagued by a host of data-related pathologies, including noisy or corrupted signals, partial or missing modalities, semantic misalignment between data streams, and the propagation of societal biases. Such deficiencies can silently subvert model performance, leading to unreliable recommendations and eroding user trust. The First International Workshop on Data Quality-Aware Multimodal Recommendation (DaQuaMRec) is convened to establish a dedicated, international forum to confront these fundamental challenges. Our objective is to drive research into new frameworks for diagnosing, measuring, and addressing data quality issues in multimodal recommendations. By focusing on data rather than just model architecture, DaQuaMRec seeks to develop more robust, equitable, and reliable recommender systems, prioritizing data quality in research.
Citation
Claudio Pomo,
Dietmar Jannach,
Yubin Kim,
Daniele Malitesta,
Alberto Carlo Maria Manchio,
Julian McAuley,
Alessandro Melchiorre,
Nawaz
First International Workshop on Data Quality-Aware Multimodal Recommendation (DaQuaMRec)
RecSys '25: Proceedings of the Nineteenth ACM Conference on Recommender Syste, 2025.
BibTeX
@inproceedings{ClaudioPomo2025DaQuaMRec,
title = {First International Workshop on Data Quality-Aware Multimodal Recommendation (DaQuaMRec)},
author = {Claudio Pomo and Dietmar Jannach and Yubin Kim and Daniele Malitesta and Alberto Carlo Maria Manchio and Julian McAuley and Alessandro Melchiorre and Nawaz},
booktitle = {RecSys '25: Proceedings of the Nineteenth ACM Conference on Recommender Syste},
year = {2025}
}