Proton: Probing Schema Linking Information from Pre-trained Language Models for Text-to-SQL Parsing

TL;DR

This paper introduces a novel unsupervised probing framework using Poincaré distance to extract relational structures from pre-trained language models, enhancing schema linking for text-to-SQL parsing and achieving state-of-the-art results.

Contribution

The work presents a new unsupervised probing method to extract relational structures from PLMs, improving schema linking in text-to-SQL parsing without additional parameters.

Findings

Probing relations capture semantic correspondences more robustly than rule-based methods.

The framework achieves new state-of-the-art performance on three benchmarks.

Qualitative analysis confirms the effectiveness of the probing procedure.

Abstract

The importance of building text-to-SQL parsers which can be applied to new databases has long been acknowledged, and a critical step to achieve this goal is schema linking, i.e., properly recognizing mentions of unseen columns or tables when generating SQLs. In this work, we propose a novel framework to elicit relational structures from large-scale pre-trained language models (PLMs) via a probing procedure based on Poincar\'e distance metric, and use the induced relations to augment current graph-based parsers for better schema linking. Compared with commonly-used rule-based methods for schema linking, we found that probing relations can robustly capture semantic correspondences, even when surface forms of mentions and entities differ. Moreover, our probing procedure is entirely unsupervised and requires no additional parameters. Extensive experiments show that our framework sets new state-of-the-art performance on three benchmarks. We empirically verify that our probing procedure can indeed find desired relational structures through qualitative analysis. Our code can be found at https://github.com/AlibabaResearch/DAMO-ConvAI.