Fig. 3: Skill valuation concerning different job contests.

a We calculated the influence of level lv as \({r}_{s}^{{\rm{lv}}}=\frac{{\sum }_{i,j}{\mathbb{1}}\{{{\rm{lv}}}_{j}^{(i)}={\rm{lv}}\}({v}_{j}^{(i)}-{v}_{{s}_{j}^{(i)}})/{v}_{{s}_{j}^{(i)}}}{{\sum }_{i,j}{\mathbb{1}}\{{{\rm{lv}}}_{j}^{(i)}={\rm{lv}}\}}\), where and \({v}_{{s}_{j}^{(i)}}\) denotes the averaged value of skill \({s}_{j}^{(i)}\). We also show the 95% confidence interval (CI) in the figure, where data are presented as mean values ± CI. We use different colors to indicate level influence on different bounds. b CSVN assigns the skills with temporal embeddings to catch their dynamic changes, we show the average value of some skills at different time intervals. The shadow shows the 95% confidence interval, where data are presented as mean values ± CI. We use different colors to indicate different skills. c The value of some randomly selected skills with different length of working experience. The shadow shows the 95% confidence interval, where data are presented as mean values ± CI. We use different colors to indicate different skills. d To analyze the value of skills with respect to different companies, we draw the value distribution of some popular skills in five famous Chinese Internet companies on boxplots. The box shows the quartiles of the dataset. The whiskers extend to show the rest of the distribution except for outliers. Specifically, as a common practice, we regarded the samples outside 1.5 times interquartile range (IQR) above the upper quartile or below the lower quartile as outliers. We use different colors to indicate different skills.