Event detection in social networks

Abstract

Micro-blog services such as Twitter generate a large amount of messages carrying event information and users' opinions over a wide range of topics. The events discussed on social networks can be associated with topics, locations, and time periods. The events can be a variety, such as celebrities or political affairs, local social events, accidents, protests, or natural disasters. Messages are posted by users after they have experienced or witnessed the events happening in the real-world and they want to share their experiences immediately. People also express themselves spontaneously with respect to the social events in their social networks. Alternatively, policy-makers may want to know the feelings of users for a particular event to make informed decisions. With the increasing number of real-world events that are originated and discussed over social networks, event detection and tracking is becoming a compelling research issue. However, the traditional approaches to event detection and event tracking on large text streams are not applicable because of the following problems. First, they are not designed to deal with a large number of short and noisy messages. Second, social networks contain network structures such as friends, followers, replies, and re-tweets. Third, social network messages are associated with locations, which can be either senders' current locations or event locations. Fourth, each message is also associated with a timestamp. Messages often contain revealing and timely event information however, traditional text processing approaches assume documents are non-temporal. Moreover, given a particular time frame and a location the user is interested in, events that occurred in the given time frame from the chosen area are more valuable than others. Finding localized events has not been well studied yet.The goals of this thesis are to: (1) identify subsistent problems and challenges in event detection and tracking in streaming micro-blog text, (2) design approaches for event detection and event tracking in social networks, (3) design approaches for sentiment analysis for given event topics, and (4) evaluate the proposed approaches in real-world streaming datasets.In this thesis, our research is considered in three parts. Firstly, in order to detect emerging events from a large number of short and noisy messages, we propose an approach for the early detection of emerging hotspot events in social networks with location sensitivity. An algorithm is designed for slang conversion, synonym expansion and conceptual similarity to provide a rich semantic context for measuring message similarity to improve clustering results. We consider the message-mentioned locations for identifying the locations of events. In our approach, we identify strong correlations between user locations and event locations in detecting the hotspot emerging events. A sliding window manager is used to keep track of messages arriving in the system. The size of the sliding window is defined as the time interval. We evaluate our...