Efficient Location-Based Event Detection in Social Text Streams

Abstract

Social networks provide a wealth of online sources about real-world events. Due to the large volume of data in social streams, the event detection suffers from high computational complexity. In this work, we present a location-based event detection approach using Locality-Sensitive Hashing to accelerate the similarity comparison. We use this approach to detect real-world events from Sina Weibo by clustering microblogs with high similarities. We propose a message-mentioned location extraction method based on the textual content based on Part-of-Speech tagging and a Support Vector Machine classifier and a novel similarity measurement considering content, location, and time between messages to improve the precision of event detection. We compare our approach with the state-of-the-art baselines on event detection, and demonstrate the effectiveness of our approach.