Improved Personalized Headline Generation via Denoising Fake Interests from Implicit Feedback

2025-08-10

AI Summary

Existing personalized news headline generation methods are highly susceptible to noise in user click streams—such as accidental clicks or abnormally short dwell times—leading to distorted user interest modeling. To address this, we propose a two-stage denoising framework: (1) a noise-aware click detection stage that identifies spurious interactions using dwell time and anomalous click surges; and (2) a multi-level temporal fusion module that dynamically models the evolution of genuine user interests. Our approach effectively disentangles false feedback from authentic preferences. Evaluated on DT-PENS—a high-quality, self-constructed benchmark dataset—we achieve significant improvements in headline generation quality, consistently outperforming state-of-the-art methods across key metrics (BLEU-4, ROUGE-L, and human evaluation). This work is the first to systematically characterize the detrimental impact of click noise on generative performance and demonstrate its mitigability through principled denoising.

Technology Category

• Natural Language Processing: Generation
• Data Mining & Knowledge Management: Intelligent Query Processing
• Machine Learning: Learning Preferences or Rankings

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Content-based information diffusion
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Accurate personalized headline generation hinges on precisely capturing user interests from historical behaviors. However, existing methods neglect personalized-irrelevant click noise in entire historical clickstreams, which may lead to hallucinated headlines that deviate from genuine user preferences. In this paper, we reveal the detrimental impact of click noise on personalized generation quality through rigorous analysis in both user and news dimensions. Based on these insights, we propose a novel Personalized Headline Generation framework via Denoising Fake Interests from Implicit Feedback (PHG-DIF). PHG-DIF first employs dual-stage filtering to effectively remove clickstream noise, identified by short dwell times and abnormal click bursts, and then leverages multi-level temporal fusion to dynamically model users' evolving and multi-faceted interests for precise profiling. Moreover, we release DT-PENS, a new benchmark dataset comprising the click behavior of 1,000 carefully curated users and nearly 10,000 annotated personalized headlines with historical dwell time annotations. Extensive experiments demonstrate that PHG-DIF substantially mitigates the adverse effects of click noise and significantly improves headline quality, achieving state-of-the-art (SOTA) results on DT-PENS. Our framework implementation and dataset are available at https://github.com/liukejin-up/PHG-DIF.

Problem

• Eliminating click noise in personalized headline generation
• Modeling evolving user interests accurately
• Improving headline relevance to user preferences

Innovation

• Dual-stage filtering removes clickstream noise
• Multi-level temporal fusion models user interests
• New benchmark dataset DT-PENS released