Abstract
Social media is a popular platform for daily communication. It is the fastest medium to get real-time information about any event. Event identification and finding relations between them is important for information retrieval, which can be useful in many situations. For example, in case of disaster management this information can be helpful in better planning of response operations for future events. However, discovering the important events from a social media data is a challenging task due to the sheer volume of data. In this paper, we present an automated approach for discovering events and their relationships from Twitter feeds. Our proposed approach uses a two-level clustering approach. The first level clustering identifies major events among diverse tweets, and the second level clustering identifies sub-events of a given major event by considering their spatio-temporal and semantic relationships. We evaluate our approach on a dataset taken from twitter. Results show that the two level clustering could discover major events and associated sub-events with reasonable accuracy. We also discuss the implications of the automated approach of event discovery in emergency planning and emergency response evaluation.