Transition-based parsing is a method in computational linguistics that incrementally builds syntactic structures, using a sequence of predefined actions or transitions. It is efficient for dependency parsing, leveraging machine learning to predict the next action based on the current state of the parse.