The prevalence of Web 2.0 systems leads to a huge growth of information and intensifies the competition for user attention. Hence understanding what make one news article popular and how to predict its popularity has attracted a lot of interest. In this paper, we focus on the cold-start problem of popularity prediction, where only content features are used as input. A hierarchical neural network is proposed to build distributed representation of news articles. This model could capture specific n-grams and model the sequenced relations among sentences. Experiments on the real data from an online news portal demonstrate that the proposed method outperforms traditional methods and is especially good at identifying underlying popular news.
Paper
Full text
Hierarchical neural network for online news popularity prediction
Semantic Scholar · Computer Science · 2017
Abstract
The prevalence of Web 2.0 systems leads to a huge growth of information and intensifies the competition for user attention. Hence understanding what make one news article popular and how to predict its popularity has attracted a lot of interest. In this paper, we focus on the cold-start problem of popularity prediction, where only content features are used as input. A hierarchical neural network is proposed to build distributed representation of news articles. This model could capture specific n-grams and model the sequenced relations among sentences. Experiments on the real data from an online news portal demonstrate that the proposed method outperforms traditional methods and is especially good at identifying underlying popular news.