A Deep Learning Framework for Visual Attention Prediction and Analysis of News Interfaces

2025-03-21

AI Summary

To address the limited scale and insufficient demographic representativeness of existing datasets in news interface saliency prediction, this work introduces the first cross-age visual attention prediction framework specifically designed for news UIs. Methodologically, we propose a novel three-stage optimization pipeline—salience map generation, grid-based scoring, and normalization—by integrating DeepGaze IIE and SaRa models for the first time. Validation is conducted via an eye-tracking and mouse-tracking study involving 405 participants aged 13–70. Key contributions include: (1) a 10.7% improvement in SOR performance; (2) empirical validation that mouse tracking reliably approximates eye-movement behavior (sAUC = 0.86); (3) statistically significant age effects on text–image attention allocation (p < 0.05, η² > 0.01): users aged 36–70 exhibit stronger textual bias, whereas those aged 13–35 attend more to images; and (4) demonstration that large-scale mouse-tracking data effectively supports saliency modeling.

Technology Category

• Humans and AI: Intelligent User Interfaces
• Computer Vision: Motion & Tracking
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements

Abstract

News outlets' competition for attention in news interfaces has highlighted the need for demographically-aware saliency prediction models. Despite recent advancements in saliency detection applied to user interfaces (UI), existing datasets are limited in size and demographic representation. We present a deep learning framework that enhances the SaRa (Saliency Ranking) model with DeepGaze IIE, improving Salient Object Ranking (SOR) performance by 10.7%. Our framework optimizes three key components: saliency map generation, grid segment scoring, and map normalization. Through a two-fold experiment using eye-tracking (30 participants) and mouse-tracking (375 participants aged 13--70), we analyze attention patterns across demographic groups. Statistical analysis reveals significant age-based variations (p<0.05, {epsilon^2} = 0.042), with older users (36--70) engaging more with textual content and younger users (13--35) interacting more with images. Mouse-tracking data closely approximates eye-tracking behavior (sAUC = 0.86) and identifies UI elements that immediately stand out, validating its use in large-scale studies. We conclude that saliency studies should prioritize gathering data from a larger, demographically representative sample and report exact demographic distributions.

Problem

• Develop demographically-aware visual attention prediction models for news interfaces
• Improve Salient Object Ranking performance using deep learning techniques
• Analyze attention patterns across different age groups in UI interactions

Innovation

• Deep learning enhances SaRa with DeepGaze IIE
• Optimizes saliency map, scoring, and normalization
• Uses eye-tracking and mouse-tracking for analysis