Time Matters: Enhancing Pre-trained News Recommendation Models with Robust User Dwell Time Injection

TL;DR

The paper tackles the challenge of noisy click signals in news recommendation by leveraging dwell time as an explicit feedback cue. It introduces two dwell-time infusion strategies, DweW (pre-injection) and DweA (post-injection), to refine user representations and integrate temporal signals with semantic news understanding via PLM-based embeddings. Empirical results on the MSN News dataset show that DweW and DweA improve $AUC$, $MRR$, and $NDCG$ metrics over strong baselines, with notable robustness when dwell time data are missing or delayed. The work demonstrates that modeling both effective clicks and the interaction between semantics and dwell time yields more accurate, industry-ready recommendations, addressing practical deployment concerns.

Abstract

Large Language Models (LLMs) have revolutionized text comprehension, leading to State-of-the-Art (SOTA) news recommendation models that utilize LLMs for in-depth news understanding. Despite this, accurately modeling user preferences remains challenging due to the inherent uncertainty of click behaviors. Techniques like multi-head attention in Transformers seek to alleviate this by capturing interactions among clicks, yet they fall short in integrating explicit feedback signals. User Dwell Time emerges as a powerful indicator, offering the potential to enhance the weak signals emanating from clicks. Nonetheless, its real-world applicability is questionable, especially when dwell time data collection is subject to delays. To bridge this gap, this paper proposes two novel and robust dwell time injection strategies, namely Dwell time Weight (DweW) and Dwell time Aware (DweA). Dwe} concentrates on refining Effective User Clicks through detailed analysis of dwell time, integrating with initial behavioral inputs to construct a more robust user preference. DweA empowers the model with awareness of dwell time information, thereby facilitating autonomous adjustment of attention values in user modeling. This enhancement sharpens the model's ability to accurately identify user preferences. In our experiment using the real-world news dataset from MSN website, we validated that our two strategies significantly improve recommendation performance, favoring high-quality news. Crucially, our approaches exhibit robustness to user dwell time information, maintaining their ability to recommend high-quality content even in extreme cases where dwell time data is entirely missing.

Figures (5)

• Figure 1: Dwell time as a explicit feedback signal to refine user historical behaviors.
• Figure 2: Dwell time distribution.
• Figure 3: The structure of PLM-NR.
• Figure 4: The overall structure of our methods.
• Figure 5: Performance Comparison of Recommendation Algorithms Across Different Filter Thresholds.