Language systems need different structures for different jobs. AMR represents the meaning of a sentence as a rooted graph; semantic parsing maps language into a formal representation; relation extraction identifies selected facts between entities. They overlap, but they are not interchangeable.
Start with the question
Use AMR when you need a rich, sentence-level account of events, roles and core meaning. Use relation extraction when the task is a bounded fact such as organisation–acquired–company. Use semantic parsing when an application needs an executable or queryable formal form.
Do not confuse meaning with action
An AMR graph can describe that a person requests an action; it should not silently become permission to execute it. Keep semantic representation, world state and workflow execution separate. This makes systems inspectable and allows each layer to fail safely.
For notation and worked examples, read AMR Graphs Explained . For the architecture boundary, see AMR and Execution Graphs .