We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

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

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 this https URL
Comments: Accepted at KDD 2022
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2206.14017 [cs.CL]
  (or arXiv:2206.14017v2 [cs.CL] for this version)

Submission history

From: Binyuan Hui [view email]
[v1] Tue, 28 Jun 2022 14:05:25 GMT (663kb,D)
[v2] Sat, 6 Aug 2022 09:49:48 GMT (693kb,D)

Link back to: arXiv, form interface, contact.