Fig 1.
(a) The graph shows the search result of a sample neologism 蜗居 wo1ju1 “living within a snail’s shell” in Google Trends with a typical sharp rise and decay pattern. (b) The snapshot of the search result for the neologism 雷 lei2 “thunder, describing a person getting shocked by something absurd”. This word’s search frequency variation cannot be included in the paradigm of the sharp rise and decay pattern. The spatial distribution of the search frequency all over the world are also provided on the web page.
Fig 2.
Four stages of neologisms’ lifetime.
A schematic graph showing the four stages of an internet neologism’s lifetime, including a rise in popularity, a retaining of popularity, a process of decay in popularity, and a subsequent retention of a low level of popularity.
Fig 3.
(a) shows the data of the 2008 neologism 非诚勿扰 fei1cheng2 wu4rao3 “Serious Suitors Only” with multiple peaks at January of 2009, May of 2010, and December of 2010 respectively. (b) gives an example of data smoothing, where the raw data and the smoothed data are indicated by the color blue and red respectively. The smoothing process helped us to pick up the main evolution tendency of the word’s popularity by eliminating the disturbances caused by data noise.
Fig 4.
The performance of the model fitting.
The SIR model fitting results are as illustrated. In the picture, the P(t) data obtained from Google Trends are grey stars. The SIR model fitting functions are denoted by the red lines. (a),(b),(c) gives the data of 蓝瘦香菇 lan2shou4xiang1gu1 “too sad to cry”, 蜗居 wo1ju1 “living within a snail’s shell/small room”, 洪荒之力 hong2huang1zhi1li4 “the force from primitive period” respectively, as well as the SIR fitting functions.
Fig 5.
The predictive power of the SIR model.
One can use the SIR model to predict the popularity evolution of a neologism based on a small amount of data before reaching the inflection point. A fraction of data to be fitted and the remaining data are denoted by red and blue circles respectively. In order to derive α and β, the testing exponential functions denoted by black squares are used to fit with the data of 舌尖上 she2jian1shang4 “on the tip of the tongue” and 低碳 di1tan4 “low-carbon” as shown in (a) and (c) respectively. By utilizing the parameters derived from this step, we can obtain the predicted popularity evolution over time, indicated by the black diamonds, according to the small amount of data, which is very close to the green fitting curves yielded by fitting the full data, as illustrated in (b), (d).