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

TL;DR

MIGA is a unified multi-task generation framework that leverages pre-trained language models to improve conversational text-to-SQL translation, achieving state-of-the-art results on benchmark datasets.

Contribution

The paper introduces a two-stage multi-task learning framework with SQL perturbations, unifying related tasks into a Seq2Seq paradigm for conversational text-to-SQL.

Findings

Achieves state-of-the-art performance on SparC and CoSQL benchmarks.

Effectively reduces error propagation through SQL perturbations.

Provides new insights into multi-task training for conversational text-to-SQL.

Abstract

Conversational text-to-SQL is designed to translate multi-turn natural language questions into their corresponding SQL queries. Most state-of-the-art 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.