An Open Dataset of Chinese Name-to-Gender Associations for Gender Prediction in Broad Scientific Research

Fig. 1 The distribution of names in the Chinese Gender dataset.  (a) prensets the distribution of names in Chinese characters and (b) presents that in Pinyin formant.  Nmaes are divided the names into ten groups according to their female ratio (ranging from 0 to1), where a higher ratio signifies a name predominantly used by women. The x-axis indicates the female ratio, while the y-axis indicates the population percentage within each group.

Fig 2 The concentration of names in Chinese characters (a) and Pinyin format (b) in the Chinese Gender dataset. The horizontal axis represents the top percent of names, and the vertical axis represents the percentage of the total population. 

Fig. 3 Name coverage in the Grantees (a) and Teenagers (b) over various frequency settings. The x-axis is the frequency of names, the y-axis is the proportion of names under a limit of name frequency. Red dotted line is located on frequency=10. Name coverage in the Grantees (c) and Teenagers (d) over various threshold settings. We show the coverage of the Grantees(c) and Teenagers(d) by setting various thresholds, with a frequency of 10. The x-axis is the gender threshold(female=1, male=0), and the y-axis denotes the proportion of males(=<threshold) and females(>=threshold). Adjusting the threshold determines the confidence level for categorizing a name as male or female, affecting the number of classified individuals. Starting at a threshold of 0.5, it can cover all predictable individuals.

Fig. 4 The relation between precision and threshold. Here we restrict the frequency of names to 10 or more and perform the analysis separately for male(a,c) and female(b,d) groups. The upper and lower sides denote the results of the Grantees(a,b) and Teenagers(c,d). The threshold range is 0.5-1, with 0.01 increment.