Aspect-based sentiment analysis is a subclass of sentiment analysis tasks, focusing on judging the sentiment tendency of entities or attributes, and has received extensive attention due to fine-grained analysis results. The current research mostly summarizes the research from a single point of aspect extraction or sentiment classification, and lacks of research on the methods and problems in aspect-based sentiment analysis from the overall perspective. By using the method of literature investigation, an in-depth overview of the typical methods and solutions is given in the field of aspect-based sentiment analysis from three aspects: aspect information extraction, aspect information sentiment classification and aspect-based sentiment analysis joint modeling. Then, we discuss some of the most challenging problems existing in the field in terms of model adaptability, error transmission, knowledge and reasoning, and proposes future research opportunities in terms of strengthening knowledge representation, multi-task learning and integrating knowledge and reasoning from the overall perspective of practical application.