An attention-based Ranking Model for social media

The document describes applying a visual attention model to help address information overload. It proposes using a guided search 2.0 model, which predicts where visual attention is drawn, to rank communications like messages and blog posts. Features that attract attention include topics of interest to the user, message sender, and abnormal levels of activity. A feature map is created to represent the saliency or attractiveness of each item for each feature. Weighting different features allows computing an overall saliency map to select the most important items given limited time and attention. The model aims to integrate concepts of limited capacity and vigilance to adapt to different contexts. Future work includes evaluating how best to combine feature importance into the saliency map.