Event Detection Based On Social Media Data

The world is witnessing a time of big data and information explosion.Every day there is huge amount of news articles and event information represented by data of different modalities and sources.In the modern life,news events are not only reported in various news websites hosting articles,newspaper,and video broadcast,but also generally discussed by people in a variety of social media platform(e.g.Twitter,Facebook,and Instagram).Using traditional natural language processing techniques we can easily detect news event and extract news relevant information from news websites.However,for a particular and specific event,different news websites may present content variously,which makes event identification a non-trivial problem;also news articles are often not in real-time and could not render the dynamic situations,it is of great significance to extract the relevant information from the social media platforms,which reflect the comprehensive understanding of the event.In this work,we propose a framework that combines the news website reports and social media platforms data,using the social media information to enhance and enrich the detected events in news articles.Our proposed news-event enhancement framework consisting a pipeline of three components: data collection,news event initialization and event enhancement.We first utilize DBSCAN clustering to detect news events from official news articles,then we apply the unofficial information details extracted from social media data(Twitter)to enhance understanding of detected events in official news articles.Specifically,we propose a novel processing framework for event enhancement,in which event shared keywords,shared social tags and shared temporal information that extracted from news clusters are used to query relevant social information from Twitter.To demonstrate the effectiveness of our framework,we crawl a large multi-source event dataset from official news website and Twitter,and display it two real-world events that drew considerable public attention.