Fig. 9

a: Gender detection rate by year. The algorithm gender determination typically succeeds reliably for articles that are published since 2007 as indicated by this example with relative constant percentages of male (55 %), female (31 %), unisex (4 %) and unknown (11 %) gender, respectively. Before 2007 the dominance of initials prevents the correct lexical gender identification by first names. b: Country-specific gender detection rate. The country-specific distribution of the algorithmic gender detection documents a clear dependence on the authors’ country with high detection rates (>80 % male or female) for the majority of the top 20 productive countries and low detection rates for the Asian states China (CN), South Korea (KR) and Taiwan (TW) that are characterized by a high rate of unisex names and India (IN) with almost 50 % of unknown names