Rabeeh Abbasi

Rabeeh Abbasi

Doing research in social media and network analytics

Detecting Unspecified Events on Twitter

Published:

Detecting Unspecified Events on Twitter by Dr Zafar Saeed (2020)

Social media is a modern paradigm where user-generated contents are published regularly. In the last decade, Twitter has become a popular platform among the available social media services for sharing opinions, experiences, news, and views in the real-time. A vast amount of real-time sharing is influenced by the events emerging in real world. It provides an exciting opportunity for detecting events happening around the world. Despite the inherent capabilities of Twitter for generating real-time information, the content (tweets) published on Twitter are short and pose diverse challenges for detecting and interpreting event-related information. Hence, detecting events from Twitter data requires complex techniques. Event detection techniques must be scalable and robust, so that they can deal with the huge volume and noise associated with Twitter data. Existing event detection methods mostly rely on burstiness, mainly the frequency of words and their co-occurrences. However, burstiness sometimes dominates other relevant details in the data which could be equally significant. Besides, the topological and temporal relationships in the data are often ignored. This research infers informational patterns about emerging events by characterizing the growth and divergence in the Twitter stream. We propose a novel graph-based approach, called the Dynamic Heartbeat Graph (DHG), which transforms the Twitter stream into temporal graphs. DHG suppresses dominating topics in the subsequent data stream, after their first detection. It measures the divergence in bursty topics and the cohesion in the topological structure of temporal graphs to detect events. The empirical evaluation of results on three well-known benchmark datasets (FA Cup, Super Tuesday, and the US Election) shows that the proposed DHG approach has superior performance and efficiency in comparison with the state-of-the-art approaches. The execution time analysis shows that the proposed approach is 47%, 64%, and 74% faster than the second-best approach used in the baselines. The proposed approach is implemented in the form of a research prototype that demonstrates the functionality and utility of an event detection system. In the end, the thesis concludes the study by reviewing the implications and accomplishments of the aims and objectives and presents future research directions.