Traditional Japanese Haiku Generator using RNN Language Model
Guanming Shao, Yosuke Kobayashi, Jay Junichi Kishigami · 2018
Recent progress in deep learning demonstrated that machines can do well in regularized tasks. However, artistic activities such as poetry generation are still widely regarded as human special abilities. We pay special attention to the traditional Japanese Haiku and show that the machine can be as good as many contemporary poets. This paper proposes a Japanese Haiku model based on recurrent neural network language model. It accepts a set of keywords as topics, generating poetry in each generation by looking at each keyword. Experimental results show that our system can produce Haiku like humans.