The datasets like WikiSQL is that the table corresponding to question is given. But in real industrial application, we have 100+ tables for 1 new question. Thank you!