Exploring Deep Learning Approaches for Text Summarization and Text Generation

In a wide range of domains, Deep Learning (DL) approaches have achieved cutting-edge outcomes. They include several processing layers that allow them to understand the hierarchical data representation. DL architectures and designs have recently emerged in the context of Natural Language Processing (NLP). The developments that have taken place in the field of deep generative modelling are briefly described in this study. Online documents have significantly increased in volume recently. Therefore, for these documents to be useful, a summary is required. This study explore the most recent DL based extractive and abstractive summarization methodologies. Additionally, ROUGE is the primary evaluation metric used to rate the effectiveness of the text summarization procedure. Using Generative Adversarial Networks (GAN), Variational Auto Encoders (VAE), and Recurrent Neural Networks (RNN), this study explores various DL models that have been applied to the generation of text. Additionally, it was found that the majority of text summarization models encountered difficulties like the lack of a precision, vocabulary, repetition, incorrect sentences, and made-up facts. This study analyses the difficulties that each approach faces during the generation and summarization process. Following the discussion of the state of art, recommendations for future study in this field are presented.