Reference

Le Chen, Chi Zhang, Christo Wilson. Tweeting under pressure: analyzing trending topics and evolving word choice on Sina Weibo. ONS’13 Proceedings of the first ACM conference on Online social networks, pp89-100. October 07-08,2013

This article examines the overall impact of censorship on discourse in social media. To be more specific, the article examines how censorship impacts discussions on Weibo, and how users adapt to avoid censorship. The authors gather tweets and comments from 280K politically active Weibo users for 44 days and use NLP techniques to identify to identify trending topics. This article firstly makes a brief introduction on Sina Weibo itself and the ways to examine the censorship. The article also reveals the background of both Sina Weibo, the Chinese governmental regulation of the web, and studies of censorship on Weibo. The methodology is also showed in this article in order to make the data convictive. Through characterizing Weibo users and examining daily activity, this article finds out that trending topics are located with word segmentation, topic extraction, manual validation, and labeling tweet and comments. This article also makes an analysis on trending topics through high-level overview, impact of censorship, and discussion.In the conclusion, the article states that though positive correlations between censorship and user engagement are observed, these relations could be even higher in the complete absence of censorship. Besides, the authors also observe a strong relationship between censorship and the use of morphs. -Li Lin

Yuling Yao, Aki Vehtari, Daniel Simpson, and Andrew Gelman. Using Stacking to Average Bayesian Predictive Distributions (with Discussion). Bayesian Analysis, Volume 13, Number 3 (2018), 917-1007. 2018

This paper discusses Average Bayesian Predictive Distributions from a Mathematic perspective. This paper first makes a brief introduction on Bayesian Average. The authors take the idea of stacking from the point estimation literature and generalize to the combination of predictive distributions. They extend the utility function to any proper scoring rule and use Pareto smoothed importance sampling to efficiently compute the required leave-one-out posterior distributions. They also compare stacking of predictive distributions to several alternatives. Based on simulations and real-data applications, this paper recommends stacking of predictive distributions, with bootstrapped-Pseudo-BMA as an approximate alternative when computation cost is an issue. -Li Lin

Borong Lyu, Xinhui Shao, Yinbo Huang and Yuyang Xie. Analysis of Topic Influence and Post Features of Sina-Weibo. Advances in Computer Science Research, volume 80, Dec 2017

This paper focused on the most popular microblog in China – Sina Weibo and discussed topic influences. Firstly, this paper decomposed topic structure into Average Fundamental Popularity and Information Tipping Points. Next, the authors collected 10455 available Information Tipping Points and obtained all the posts on each one by Python. Statistical methods such as the Spearman Rank Correlation Coefficient Method, the Levene Homogeneity of Variance Test, and One-way Analysis of Variance were used to analyze the relationship among some features. They find that eight of all the features have strong relationships with the Number of Weibo Posts. Furthermore, the total number of posts on one topic (the Number of Weibo Posts) can represent the toxic impact. This paper established a predictive model via the regression method to predict the Number of Weibo Posts. -Li Lin

Zhihu. Two ways of calculating trending topics. Dec 2017, Available at https://www.zhihu.com/question/264859637.

This is an online answer published in the question-and-answer website Zhihu. The replier introduced two basic formulas which are generally used in all treading topics algorithm. The first formula is the Bayesian Average: WR=v/(v+m)*R+m/(v+m)C. WR represents the weighted average of each topic. The higher the WR, the hotter the topic is. The main idea of this formula is putting different weights to different factors. The second formula is Newton’s Law of Cooling Formula: The rate of loss of heat by a body is directly proportional to the temperature difference between system and surroundings, provided the difference is small. The main purpose of the use of this formula is making sure the newest topics can rise to the top position of the treading list.-Rujin Guo

Kathy Lee, Diana Palsetia, Ramanathan Narayanan, etc. Twitter trending Topic Classification. 2011 11th IEEE International Conference on Data Mining Workshop. 2011. Available at: http://cucis.ece.northwestern.edu/publications/pdf/LeePal11.pdf

This article mainly talks about how Twitter does its trending topic classification. The authors experimented with two approaches for topic classification: 1) the well-known Bag-of-Words approach for text classification 2) network-based classification. In the text-based classification method, they construct word vectors with trending topic definition and tweets, and the commonly used weights are used to identify the topics. In the network-based classification method, they identify the top five similar topics for a given topic based on the number of common influential users. The categories of similar topics and the number of common influential users between the given topic and its similar topics are used to classify the given topic using a decision tree learner. Because Weibo and Twitter are kinda similar to each other, we use this article to analogy inference the working process of Weibo.-Rujin Guo

Lei, K., Liu, Y., Zhong, S., Liu, Y., Xu, K., Shen, Y., & Yang, M. (2018). Understanding User Behavior in Sina Weibo Online Social Network: A Community Approach. IEEE Access, 6, 13302–13316. https://doi.org/10.1109/ACCESS.2018.2808158

This article focused on the characters of Weibo user behaviors in tweeting, retweeting, and commenting. The authors built a Weibo community graph to analyze user behaviors, clustering the coefficients of the community graph and exploring the impact of user popularity on social network sites. Bipartite graphs and one-mode projections are used to analyze the similarity of retweeting and commenting activities, which reveal the weak correlations between these two behaviors. In addition, to characterize the user retweeting behaviors deeply, the authors also studied the tweeting and retweeting behaviors in terms of the gender of users. They observed that females are more likely to retweet than males. Then they introduced an information-theoretical measure based on the concept of entropy to analyze the temporal tweeting behaviors of users. Finally, the authors apply a clustering algorithm to divide users into different groups based on their tweeting behaviors, which can improve the design of plenty of applications, such as recommendation systems. This article helps us a lot to understand the basic design and user features of Weibo. We also got much inspiration from this article about how to draw our social-technical system graph.-Rujin Guo

Miao, Z., Chen, K., Fang, Y., He, J., Zhou, Y., Zhang, W., & Zha, H. (2017). Cost-Effective Online Trending Topic Detection and Popularity Prediction in Microblogging. ACM Transactions on Information Systems (TOIS), 35(3), 1–36. https://doi.org/10.1145/3001833

In this article, the authors mainly proposed a system that can detect and use the specific posts among Trending topics efficiently and make prediction of their future popularity. The system or framework was proposed to serve for third parties for academic and business use to avoid the limitation and requirements created by the microblogging companies, their massive dataset and API (Application Program Interface) systems. The article began with introducing the topic trends on microblogging services such as Twitter and the procedures about how to update the list in real time, which was beneficial to our research topic. Wenjie

Chen, Y., Li, Z., Nie, L., Hu, X., Wang, X., Chua, T., & Zhang, X. (2012). A Semi-Supervised Bayesian Network Model for Microblog Topic Classification. COLING.

Authors discovered the difficulties for microblogging applications to detect the long messages to label those microblog topics, which limited the users to explore and locate the pertinent topics or messages they want. Based on the question, the article gave a new “semi-supervised learning scheme” to detect the long message with the help of the resources distilled from external online websites. This article helped us with understanding more about Sina Weibo trending topics and its way of detecting long messages, though we finally did not de-blackbox this function. Besides, it shed lights on how Bayesian Formula worked, as we understood that each letter represented various parameters which can make different effects for the whole process. Wenjie

Harwit, E. (2014). The Rise and Influence of Weibo (Microblogs) in China. Asian Survey, 54(6), 1059–1087. https://doi.org/10.1525/as.2014.54.6.1059

This article focused on Weibo’s social impact. As Weibo absorbing a plethora of users in China as well as abroad, the articles examined the rise and influence of Weibo from various perspectives, including its political, social and commercial impacts. Besides, it also probed into the way that Chinese government controls the Weibo with “key-word censorship, high-profile punishments and arrests, and citizen self-censorship”. The article not only let us better understand Weibo’s function and influence in its socio-technical system but also provided us more information on the role of censorship in Weibo filtering process. Wenjie

Yao, W., Jiao, P., Wang, W., & Sun, Y. (2019). Understanding human reposting patterns on Sina Weibo from a global perspective. Physica A, 518, 374–383. https://doi-org.proxy.library.georgetown.edu/10.1016/j.physa.2018.11.043

In the article, the authors examined the reposting patterns and characteristics on social network sites from a more macroscopic angle. Through analyzing 1.5 billion reposting records from Sina Weibo over 15 months, they get to know the “topological properties” of users’ reposting behavior. There are three main findings: 1) the time of being posted followed power law decays, and the time of reposting is related closely to users’ stable relationships; 2) the frequency of reposting between two users obeyed “a distribution of exponentially truncated power law”; 3) basically, since most social networks use assortative methods, users’ reposting behavior shows the features of aggregation, and is highly related to user attributes. Yuting Xia

WEN WU, CLARK, M., BOMI KANG, & FINE, M. (2016). The Use of Sina Weibo and Twitter by International Luxury Hotels. Tourism Culture & Communication, 16(3), 137–145. https://doi-org.proxy.library.georgetown.edu/10.3727/109830416X14750895902837

Nowadays, microblogging sites such as Weibo and Twitter have allowed consumers to make comments on brands effectively, and it is also a good marketing strategy. The authors put emphasis on luxury hotels to see if they act differently on posting their information and leading certain discussion. It shows that luxury hotels will use a more traditional promotion post on Weibo while conversational one in Twitter, which may due to culture difference. They also points out that Twitter and Sina Weibo are increasing in use by hotel industry to build brand images and do the promotions, showing that social media such as Weibo works as an important bridge between companies and consumers. Yuting Xia

Kim, S.-E., Lee, K. Y., Shin, S. I., & Yang, S.-B. (2017). Effects of tourism information quality in social media on destination image formation: The case of Sina Weibo.Information & Management,54(6), 687–702. https://doi-org.proxy.library.georgetown.edu/10.1016/j.im.2017.02.009

The authors claims that the relationship between tourism information quality in social media and their corresponded destination image formation is worthy of examining because the former can influence the cognitive and affective images that lead to the formation of conative images related to the destination. To put a empirical study, the authors carried out a survey among Weibo users who follow official account of Gyeonggi Tourism Organization on Weibo, and find out that the two factors mentioned above has got positive relationship. This reflects the significance of social media website like Weibo has on consumer insights. Yuting Xia