Characterizing and modeling the dynamics of online knowledge-sharing networks

Abstract

Online knowledge-sharing networks, such as wikis and question-answering (Q&A) portals, present users with a channel for seeking and discussing information on diverse subjects pertaining to their interests and expertise. Through their collaborative efforts in providing guidance, raising and answering questions, and otherwise engaging in topical and technical discussions, users form interactive communities around topics of mutual interest. Analyzing the dynamics of these topic-based communities is crucial to understanding the network's inner processes, such as the flow of users across communities and the diffusion of information within it. In this thesis, we study how users relate to topic-based communities and how this relationship shapes long-term community dynamics. Using a large dataset collected from Stack Overflow, a popular programming-oriented Q&A site, we investigate several factors related to the evolution of communities in the site, including the impact of user revisits, continued activity and migration on community sustainability. Our findings motivate the development of a community evolution model based on user activity, which incorporates key aspects of community dynamics. In addition to describing the activity levels of communities over time, our new model also provides additional insight into the effects that related communities may have on one another. We expand on this insight to show how information about inter-community relationships can be used to identify macro-communities based on the dynamic flow of users.