Compared with some Chinese sentiment analysis data sets, the mainstream algorithms have important improvements in test_acc, test_loss, F1, etc. There need to be three or more innovation points in this essay(or paper). Don′t use ″positive″,″negetive″ this kind of Binary classification,use Multi classification, like ″ Fear ″,″Disgust″, ″Optimism″, ″Surprise″, ″Gratitude ,″Sadness″,″ Anger″