Abstract:Natural language is usually understood from discourse perspective. With the success of research at the lexical and sentence levels, the focus of natural language processing research has shifted to the discourse level. The object of discourse analysis is to analyze the text structure and the semantic relationship between discourse units, and thus understand the text. According to different purposes, discourse units and their relationships can be expressed as different textual structures, and the study of them can provide different levels of text comprehension. Currently, there are few studies on the inherent laws of Chinese texts and the lack of theory and method system for effective analysis and in-depth understanding of Chinese discourse has seriously restricts the relevant research and application. This study focuses on two basic features of a text, namely cohesion and coherence. From three aspects of theoretical research, resource construction, and computational model of discourse analysis, it explores the rhetorical structure (reflecting text coherence) and topic structure (reflecting text cohesion) respectively. It summarizes the current research, and presents the main problems and research trends.