Yukie Sano

Empirical analysis of collective human behavior for extraordinary events in the blogosphere.

To uncover an underlying mechanism of collective human dynamics, we survey more than 1.8 billion blog entries and observe the statistical properties of word appearances. We focus on words that show dynamic growth and decay with a tendency to diverge on a certain day. After careful pretreatment and the use of a fitting method, we found power laws generally approximate the functional forms of growth and decay with various exponents values between -0.1 and -2.5. We also observe news words whose frequencies increase suddenly and decay following power laws. In order to explain these dynamics, we propose a simple model of posting blogs involving a keyword, and its validity is checked directly from the data. The model suggests that bloggers are not only responding to the latest number of blogs but also suffering deadline pressure from the divergence day. Our empirical results can be used for predicting the number of blogs in advance and for estimating the period to return to the normal fluctuation level.

Rumor Diffusion and Convergence during the 3.11 Earthquake: A Twitter Case Study

We focus on Internet rumors and present an empirical analysis and simulation results of their diffusion and convergence during emergencies. In particular, we study one rumor that appeared in the immediate aftermath of the Great East Japan Earthquake on March 11, 2011, which later turned out to be misinformation. By investigating whole Japanese tweets that were sent one week after the quake, we show that one correction tweet, which originated from a city hall account, diffused enormously. We also demonstrate a stochastic agent-based model, which is inspired by contagion model of epidemics SIR, can reproduce observed rumor dynamics. Our model can estimate the rumor infection rate as well as the number of people who still believe in the rumor that cannot be observed directly. For applications, rumor diffusion sizes can be estimated in various scenarios by combining our model with the real data.

Zipf's Law and Heaps' Law Can Predict the Size of Potential Words

We confirm Zipf’s law and Heaps’ law using various types of documents such as literary works, blogs, and computer programs. Independent of the document type, the exponents of Zipf’ law are estimated to be approximately 1, whereas Heaps’ exponents appear to be dependent on the observation size, and the estimated values are scattered around 0.5. By definition, randomly shuffled documents reproduce Zipf’s law and Heaps’ law. However, artificially generated documents using the empirically observed Zipf’s law and number of distinct words do not reproduce Heaps’ law. We demonstrate that Heaps’ law holds for artificial documents in which a certain number of distinct words are added to empirically observed distinct words. This suggests that the number of potential distinct words considered in the creation of a given document can be predicted.

Emotional Changes in Japanese Blog Space Resulting from the 3.11 Earthquake

We quantified the emotional changes observed in social media after major disasters, focusing especially on the Japanese blog space after the Great East Japan Earthquake in 2011. We checked the appearances of Japanese adjectives and found that special emotion adjectives, such as ‘impatient’, and ‘frustrating’ which involve the want to help others but the person has no means to and feels frustrated, occur with considerably increasing frequency. To visualize social mood, we drew a co-occurrence network of adjectives showing a major topological change at the site of the quake. Measuring emotional changes after an emergency has been difficult, but, our research has the potential to achieve it.

Statistical properties of fluctuations of time series representing the appearance of words in nationwide blog data and their applications: An example of observations and the modelling of fluctuation scalings of nonstationary time series.

To elucidate the nontrivial empirical statistical properties of fluctuations of a typical nonsteady time series representing the appearance of words in blogs, we investigated approximately 3×10^{9} Japanese blog articles over a period of six years and analyze some corresponding mathematical models. First, we introduce a solvable nonsteady extension of the random diffusion model, which can be deduced by modeling the behavior of heterogeneous random bloggers. Next, we deduce theoretical expressions for both the temporal and ensemble fluctuation scalings of this model, and demonstrate that these expressions can reproduce all empirical scalings over eight orders of magnitude. Furthermore, we show that the model can reproduce other statistical properties of time series representing the appearance of words in blogs, such as functional forms of the probability density and correlations in the total number of blogs. As an application, we quantify the abnormality of special nationwide events by measuring the fluctuation scalings of 1771 basic adjectives.

Fluctuation scaling in online social media

To examine number fluctuations in online social media, we focus on the daily appearances of adjectives, abdominal particles, and conjunctions in a Japanese blog space. We show that a non-trivial fluctuation scaling law exists between the standard deviation and the mean. When the number of appearances becomes larger, the standard deviation linearly increases. We introduce a random posting (RP) model, which is inspired by the random diffusion model, to reproduce the observed scaling law. We confirm that the RP model can estimate number fluctuations in a Japanese blog space from both analytical and simulated results.