Event Detection from Social Media: 5W1H Analysis on Big Data

Increasing number of social media content shared across the globe in real time creates a fascinating area of research. For most events, social media users act as collective sensors by sharing important information about events of any scale. Due to its real time nature, social media is much quicker to respond to such events relative to conventional news media. This paper proposes an event detection system which provides 5W1H (What, Where, When, Why, Who, How) analysis for each detected event. We make use of a myriad of techniques such as anomaly detection, named entity recognition, automatic summary generation, user link analysis. Our experimental results for the system indicate a faster event detection performance compared to conventional news sources. Event analysis results are also in line with the corresponding news articles about detected events.