MIGA: A Unified Multi-Task Generation Framework for Conversational Text-to-SQL


  • Yingwen Fu Guangdong University of Foreign Studies, Guangzhou, China NetEase Games AI Lab, Guangzhou, China
  • Wenjie Ou NetEase Games AI Lab, Guangzhou, China
  • Zhou Yu Columbia University
  • Yue Lin NetEase Games AI Lab, Guangzhou, China




SNLP: Generation, ML: Transfer, Domain Adaptation, Multi-Task Learning, SNLP: Conversational AI/Dialogue Systems, SNLP: Language Models


Conversational text-to-SQL is designed to translate multi-turn natural language questions into their corresponding SQL queries. Most advanced conversational text-to-SQL methods are incompatible with generative pre-trained language models (PLMs), such as T5. In this paper, we present a two-stage unified MultI-task Generation frAmework (MIGA) that leverages PLMs’ ability to tackle conversational text-to-SQL. In the pre-training stage, MIGA first decomposes the main task into several related sub-tasks and then unifies them into the same sequence-to-sequence (Seq2Seq) paradigm with task-specific natural language prompts to boost the main task from multi-task training. Later in the fine-tuning stage, we propose four SQL perturbations to alleviate the error propagation problem. MIGA tends to achieve state-of-the-art performance on two benchmarks (SparC and CoSQL). We also provide extensive analyses and discussions to shed light on some new perspectives for conversational text-to-SQL.




How to Cite

Fu, Y., Ou, W., Yu, Z., & Lin, Y. (2023). MIGA: A Unified Multi-Task Generation Framework for Conversational Text-to-SQL. Proceedings of the AAAI Conference on Artificial Intelligence, 37(11), 12790-12798. https://doi.org/10.1609/aaai.v37i11.26504



AAAI Technical Track on Speech & Natural Language Processing