Social Synchrony in Online Social Networks and its Application in Event Detection from Twitter Data

Abstract

Online communication is fast becoming an integral part of how our society functions. As a result, studying the patterns of online collective behaviors - especially over online social networks - is important. This work has two parts firstly, redefining an online collective phenomenon called social synchrony and secondly, applying the concept of online social synchrony on the activities over online social networks for event detection. Social synchrony is a particular kind of collective social behavior where the number of people who perform a certain action first increases and then decreases. We redefine this phenomenon and propose a method to detect it. Then we propose a method to detect the presence of events from Twitter data using the concept of social synchrony. We implement this method on a Twitter dataset of 3.3 million tweets geotagged for India. We compared the results with manually assigned labels. We got precision of 1.0 and recall of 0.87 for detecting the presence of events. Also, we tested the generality of the hypothesis made in this method on a publicly available Twitter dataset of 1.2 million tweets that is collected from a different time period with no geotag. The results indicate that the hypothesis is applicable to this dataset as well, hinting at its possible generality.