(a)
(b)
(c)
Figure 8: We present prediction results of Chinese word in (a). We also show some examples of synthetic datasets with unbalance ratios and in (b) and (c), respectively.