Please note that registration for the event closes at 2:00pm CET on Dec 12. The meeting link will be sent to all registered attendees shortly after 2:00pm.
Members of the Dissertation Committee:
Chair: Gerardo Iñiguez (CEU DNDS, voting)
Supervisor: János Kertész (CEU DNDS, non-voting)
Member: Petra Kralj Novak (CEU DNDS, voting)
Reviewer: Daniel Romero (University of Michigan, voting)
Reviewer: Zoltán Kmetty (Eötvös Loránd University, voting)
ABSTRACT / In this thesis, I investigate attention dynamics on the Chinese microblogging site Sina Weibo by studying the popular hashtags on the real-time Hot Search List (HSL). The objective goal of my thesis is to provide insights into the emergence mechanisms and popularity dynamics of hashtags on the one hand, and the effects of interventions by the microblog service providers on the other hand. Furthermore, I analyze the consequences of a strong external effect with long lasting impact on the popularity ranking list on the example of COVID-19. To achieve that, I investigate the evolution dynamics of the repost network of popular hashtags in the phase of their emergence, the ranking dynamics after the successful hashtags reach the system-wide level of popularity, and identify anomalous patterns that can be attributed to intentional measures taken by the service provider deviating from the principles of an objective ranking.
Firstly, I study the network dynamics in the prehistory of successful hashtags before they become popular. I show that the time of the day when the hashtag is born has an impact on the time needed to get to the HSL. Analyzing this time I distinguish two extreme categories which I label: a) "Born in Rome", which means hashtags are mostly first created by super-hubs or reach super-hubs at an early stage during their propagation and thus have the chance to gain immediate wide attention from the broad public, and b) "Sleeping Beauty", meaning the hashtags gain little attention at the beginning and reach system-wide popularity after a considerable time lag. The evolution of the repost networks of successful hashtags before getting to the HSL shows two types of growth patterns: "smooth" and "stepwise". The former is usually dominated by a super-hub and the latter results from consecutive waves of contributions of smaller hubs. The repost networks of unsuccessful hashtags exhibit a simple pattern of evolution.
Secondly, I study the dynamics of hashtags after they appear on the HSL, including rank trajectory clustering, duration, and ranking dynamics. I characterize the rank dynamics by the time spent by hashtags on the list, the time of the day they appear there, the rank diversity, and by the ranking trajectories. I show how the circadian rhythm affects the popularity of hashtags, and observe categories of their rank trajectories by a machine learning classification algorithm. By analyzing patterns of the ranking dynamics I identify anomalies that are likely to result from the platform provider’s intervention into the ranking, including the anchoring of hashtags to certain ranks on the HSL. I propose a simple model of ranking that explains the mechanism of this anchoring effect.
Thirdly, I present the effect of COVID-19 on Weibo's public attention dynamics by studying the correlations between different content categories of hashtags with real-world COVID situations in different periods of the pandemic. I show how the specific events, measures and developments during the epidemic affected the emergence of different kinds of hashtags and the ranking on the HSL. A dramatic increase of COVID-19 related hashtags started to occur on HSL when the transmission of the disease between humans was announced, and soon COVID-related hashtags occupied 30-70% HSL, however, with content changing according to the social and health related events. I give an analysis of how the hashtag topics changed during the investigated time span and conclude that there were three periods to distinguish during the time of observation. In period 1, I see strong topical correlations and clustering of hashtags; in period 2, the correlations are weakened, without a clustering pattern; in period 3, I see an increased clustering but not as strong as in period 1. I further explored the dynamics of HSL by measuring the ranking dynamics and the lifetimes of hashtags on the list. This way I could obtain information about the decay of attention, which is important for decisions about the temporal placement of governmental measures to achieve permanent awareness. Furthermore, I find an abnormally higher rank diversity in the top 15 ranks on HSL due to the COVID-19 related hashtags, indicating the possibility of intervention from the platform provider.
BIO / Hao is interested in building mathematical models to solve real world problems. Since undergraduate studies, she has gained mathematical modeling experiences during participation in the Mathematical Contest in Modeling (MCM) and the summer camp in Chinese Academy of Sciences. During her MSc studies, many of her research and project experiences were in an area closely aligned with machine learning of high-dimensional data. Hao's Master's thesis was on learning Bayesian network structure from data. Currently, Hao is interested in understanding hot topics propagation, popularity prediction and ranking in social media network such as Sina Weibo, the “Twitter of China”.