In this paper, we introduce a novel methodology to efficiently construct a\ncorpus for question answering over structured data. For this, we introduce an\nintermediate representation that is based on the logical query plan in a\ndatabase called Operation Trees (OT). This representation allows us to invert\nthe annotation process without losing flexibility in the types of queries that\nwe generate. Furthermore, it allows for fine-grained alignment of query tokens\nto OT operations. In our method, we randomly generate OTs from a context-free\ngrammar. Afterwards, annotators have to write the appropriate natural language\nquestion that is represented by the OT. Finally, the annotators assign the\ntokens to the OT operations. We apply the method to create a new corpus OTTA\n(Operation Trees and Token Assignment), a large semantic parsing corpus for\nevaluating natural language interfaces to databases. We compare OTTA to Spider\nand LC-QuaD 2.0 and show that our methodology more than triples the annotation\nspeed while maintaining the complexity of the queries. Finally, we train a\nstate-of-the-art semantic parsing model on our data and show that our corpus is\na challenging dataset and that the token alignment can be leveraged to increase\nthe performance significantly.\n