CrowdStory: Multi-layered event storyline generation with mobile crowdsourced data

Online social media, like Twitter, can be viewed as a kind of source of mobile crowdsourced data. It can be utilized to understand real-world social events. Recently, many researchers have worked on event summarization from this kind of data. However, those works fail to present the dynamic evolution of the event and characterize the rich correlations among posts. To this end, we propose a multilayered model for social event characterization. Based on this model, an evolutionary and multi-clue-based approach is adopted for fine-grained event summary, and a progressively multi-layer fusion approach is utilized to generate the storyline to summarize the event. Experiments show the effectiveness of our approach.