Robust Social Event Detection via Deep Clustering

Social networks are quickly becoming the primary medium for discussing what is happening around real-world events. However, it is still a challenge to detect events on social media due to its real-time nature, scale and amount of unstructured data generated. In this paper, we present a novel real-time system for detecting surrounding real-world events. Our proposed framework consists of four main components, including text filtering, text representation, deep clustering, and event merging. After filtering non-event messages, we use entities and words to represent messages. Based on text representation, we propose a novel density clustering algorithm for online event detection. The resulted sub-events are further merged based on time information and keyword similarity. Experiments on standard and real-world datasets demonstrated the effectiveness of our proposed method.