Entity Matters in News: An Association Network-enhanced Method for News Reprint Prediction

Reprint is a fast and efficient way for news media to spread information and plays an increasingly important role in evaluating the influence of media and building a brand image. News reprint prediction is a novel research question in the news diffusion field, which predicts whether a news media will reprint a piece of news in the future. During reprinting, an association network among media and entities in the news is formed that reflects their multi-dimensional dynamic interaction. Existing research primarily focuses on integrating reprint historical records and news content while reprint prediction considering the associations among media and entities in the news remains under-researched. This work develops an entity association network-enhanced method for news reprint prediction, which adopts HIN2Vec to model the dynamic interaction and incorporates the learned embedding and news content through attention mechanism to generate entity-specific content representation. The efficacy of the proposed method is validated on real-world news reprint data. Experimental results show that the fusion of entity association network helps improve the performance of reprint prediction. This research contributes to media communication literature and has significant practical implications.