Automatic text summarization is a means of information compression that uses a computer program to automatically extract key information from a text or set of texts to form a concise summary. Due to the rapid evolution of the Internet, more information than manual processing can capture accurately. To summarize the information more accurately, conveniently, and efficiently, automatic text summarization technology in natural language processing has unique advantages. In this paper, along the main line of technical discussion, two summarization techniques, extractive and generative, are comprehensively reviewed. Firstly, we describe the evaluation of the two techniques and the commonly used Chinese and English data resources; then, through case studies, we detail the six mainstream technical approaches, including those based on reinforcement learning, information theory, pointer networks, sequence annotation, pre-training models, and joint attention mechanisms, and compare their respective strengths and limitations; finally, we summarize the current challenges of extractive and generative summarization, and provide a prospective outlook on the future trends of these techniques.