Exploring Underexplored Limitations of Cross-Domain Text-to-SQL Generalization

Paper published in EMNLP 2021: arXiv

We slightly modified three databases to insert domain knowledge. To use the Spider-DK, you should merge the database folder of Spider and Spider-DK and then use the tables.json file here instead of the Spider one.

In Spider-DK.json, type 0 means there is no domain knowledge, while type 1 to 5 represent T1 to T5 in the paper. Besides, type 34 means this example contain both T3 and T4 knowledge type, and so on.

If you use the dataset in this repo, please cite the following paper:

@misc{gan2021exploring,
      title={Exploring Underexplored Limitations of Cross-Domain Text-to-SQL Generalization}, 
      author={Yujian Gan and Xinyun Chen and Matthew Purver},
      year={2021},
      eprint={2109.05157},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}