Text generation is a sub-field of Natural Language Processing (NLP) and automatically generates informative text similar to human written text by utilizing the knowledge of computational linguistics and artificial intelligence. A deep learning model is a powerful tool in NLP tasks due to its end-to-end training and multiple-level representation learning features. The deep learning model for text generation learns the vector representation, trains the deep neural network to predict the next word in the sequence, and generates meaningful text.

  Text generation is only focused on preceding text’s semantic information in the sequence and lacks the focus on analyzing information from the entire conversation or document. Discourse analysis tackles the issue in formal text generation by analyzing overall and underlying semantic information of texts and their social and historical contexts. Discourse representation is a document-level representation that concerns linguistic relations, presupposition, and co-reference within and across sentences for encoding rich semantic information. The discourse representation aware text generation model employs a deep learning model that improves the quality and robustness of text generation by exploring rhetoric characteristics of texts.

  • Natural Language Processing (NLP) tasks capturing semantics dispersed across whole text is one of the considerable and important problems that impoverished research efforts.

  • Discourse representation structures leverage text generation for producing high-quality text from formal meaning and semantic representation.

  • Discourse representation imparts document-level representations which possess rich semantic detail in concern to rhetorical intercourse, presupposition, and co-reference within and beyond the sentences.

  • Deep learning models have been employed for text generation by accurately constituting discourse representation structures.

  • The emerging framework for discourse-aware text generation is self-critical reinforcement learning, that beneficial for capturing temporal semantics in long sentences.