This is the repository containing source code for the TAPMI 2023 journal article A Heterogeneous Graph to Abstract Syntax Tree Framework for Text-to-SQL, HG2AST hereafter. If you find it useful, please cite our work.
@ARTICLE{10194972,
author={Cao, Ruisheng and Chen, Lu and Li, Jieyu and Zhang, Hanchong and Xu, Hongshen and Zhang, Wangyou and Yu, Kai},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
title={A Heterogeneous Graph to Abstract Syntax Tree Framework for Text-to-SQL},
year={2023},
volume={45},
number={11},
pages={13796-13813},
keywords={Structured Query Language;Decoding;Databases;Syntactics;Semantics;Task analysis;Computational modeling;Abstract syntax tree;grammar-based constrained decoding;heterogeneous graph neural network;knowledge-driven natural language processing;permutation invariant problem;text -to-SQL},
doi={10.1109/TPAMI.2023.3298895}}
Note that: This work focuses on leveraging small-sized bi-directional pre-trained models (e.g., BERT, ELECTRA) and labeled training data to train a specialized, interpretable and efficient local text-to-SQL parser in low-resource scenarios, instead of chasing SOTA performances. For better results, please try LLM with in-context learning (such as DINSQL and ACTSQL), or resort to larger encoder-decoder architectures containing billion parameters (such as Picard-3B and RESDSQL-3B). Due to a shift in the author's research focus in the LLM era, the relevant code repository will no longer be maintained.
The original code structure relies on very complicated graph-based encoding and tree-based decoding modules, which are difficult to maintain and not "elegant" enough for small-sized models. When token-based LLM decoder becomes mainstream, we believe the unique advantage of symbolic HG2AST framework is its lightness, ease of training and inference efficiency.
Thus, we refine the original model structure and propose a more general grammar-based Transformer decoder, called ASTormer, as the termination of a symbolic framework based on graph encoding and tree decoding. The new code repository can be found here.
@misc{cao2023astormer,
title={ASTormer: An AST Structure-aware Transformer Decoder for Text-to-SQL},
author={Ruisheng Cao and Hanchong Zhang and Hongshen Xu and Jieyu Li and Da Ma and Lu Chen and Kai Yu},
year={2023},
eprint={2310.18662},
archivePrefix={arXiv},
primaryClass={cs.CL}
}