Methods for structured outputs are techniques used to predict complex output variables that have interdependencies and specific structures, unlike simpler outputs such as single-class labels. These methods are crucial for tasks where the relationships between output components matter, such as in sequence prediction, graph construction, or hierarchical classification. Examples include applications in natural language processing for sentence parsing and machine translation, where the output needs to adhere to grammatical structures.