Text Generation in Dialogue System via Deep Neural Network

Abstract

At present, the research on text generation focuses on the article, and there is still a large gap in the study of dialogue. At present, the task of generating sentence level has achieved good results. How to generate a complete dialogue is still challenging, and the semantic coherence involved in the dialogue is a very important issue. In view of the above problems, our research is to generate dialogue text according to dialogue elements, and construct a dialogue text generation model to realize the organization of known discrete questions and key words of answer elements into a smooth dialogue. We use the deep learning method to divide the whole dialogue text generation model into two parts. Firstly, the dialogue elements are organized into the correct dialogue process and process, and then the answer elements in the answer sentence are improved into fluent sentences. Finally, by connecting the two models, we can generate fluent multi round dialogue texts according to the dialogue elements. We test and verify the two models, and the results show that this method can organize the dialogue process, improve the dialogue answers and generate the dialogue text.