A Novel Approach for Text Generation using RNN for Language Modeling

In recent years, many research works are initiated in the domain of text generation as it is the most important and needed task in natural language processing. The text generation process has various applications including chatbots, social media, language translation, news, movie script writing, and poetry comprehension. This research study provides a detailed study on text generation with the help of SRNN using LSTM and GRU for language modeling and also to generate coherent and contextually relevant text equivalent to human creativity level. The techniques such as teacher forcing, beam search, and temperature scaling are used to explore the good quality and diversity of the generated text. This study utilizes large-scale text including news articles, books and blogs to train and evaluate RNN based language models. Evaluation metrics such as perplexity, BLEU score, human evaluation to assess the fluency, coherence and overall quality of the generated text. This research study investigates the performance of RNN-based language models based on hyperparameters and the impact on different architectural variations and training strategies. This study presents several automated text generation techniques that shorten the time required to manually type and produce lengthy texts.