Abstract: Question Generation (QG) is one of the research directions of text generation in natural language processing. In this research, a machine is given a piece of text and an answer, and the machine processes the text and outputs one or more questions related to the current text and answer. At present, the research can be applied in a wide range of fields such as education, medicine and automated question and answer. However, research has shown that current supervised learning-based question generation methods still have many shortcomings. This paper described the development of the question generation, the general solution and process of the study. The paper analyzed the current state of research, divided question generation methods into four categories, and analyzed and compared the representative model architectures in each category. This paper summarized the technical challenges faced by problem generation technology and the future directions of development.