Authors:

Abstract

Many consequences in the human past can be traced back to that one well-written, well-presented speech. Speeches grasp the power to move nations or touch hearts as long as they are well-thought-out. This is why gaining the expertise of speech giving and speech writing is something we should all intent to gain. A copy-cat bot is a model that can learn the writing and talking style of a certain person and replicate it. The main objective of this research study is to apply simple Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) Recurrent Neural Networks and Gated Recurrent Unit (GRU) in developing a speech generation system that deep learns one text and then generates new text. This research looks into the generation of English transcripts of Narendra Modi’s speeches. The generated text using LSTM and GRU models has great potential. The output resulted by RNN is less realistic and pragmatic, but its variants LSTM and GRU performed better. Though the grammatical correctness and the sentence transitions were absent in generated text of LSTM and GRU, but their output is somewhat logical as compared to RNN. LSTM and GRU performed better as it generated more realistic text and training loss is small, perplexity is small and mean probability is high compared to RNN.

Keywords: Text generation, Simple Recurrent Neural Network, Long short-term memory recurrent neural networks, Gated recurrent, Speeches, Political speech generation