This paper proposes a self-attention enhanced Recurrent Neural Networks for the task of poetry generation in Hindi language. The proposed framework uses Long Short-Term Memory with multi-head selfattention mechanism. We have utilized the multi-head self-attention component to further develop the element determination and hence protect reliance over longer lengths in the recurrent neural network architectures. The paper uses a Hindi poetry dataset to train the network to generate poetries given a set of words as input. The two LSTM models proposed in the paper are able to generate poetries with significant meaning.

Keywords : Hindi Poetry Generation, Text Generation, Poetry Generation, Long Short-Term Memory.