Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer

Tianyi Li; Sujian Li; Mark Steedman

doi:10.18653/v1/2021.iwpt-1.4

Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer

Tianyi Li, Sujian Li, Mark Steedman

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Strong and affordable in-domain data is a desirable asset when transferring trained semantic parsers to novel domains. As previous methods for semi-automatically constructing such data cannot handle the complexity of realistic SQL queries, we propose to construct SQL queries via context-dependent sampling, and introduce the concept of topic. Along with our SQL query construction method, we propose a novel pipeline of semi-automatic Text-to-SQL dataset construction that covers the broad space of SQL queries. We show that the created dataset is comparable with expert annotation along multiple dimensions, and is capable of improving domain transfer performance for SOTA semantic parsers.

Original language	English
Title of host publication	Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021)
Editors	Stephan Oepen, Kenji Sagae, Reut Tsarfaty, Gosse Bouma, Djamé Seddah, Daniel Zeman
Place of Publication	Stroudsburg, PA, United States
Publisher	Association for Computational Linguistics (ACL)
Pages	38-49
Number of pages	12
ISBN (Electronic)	978-1-954085-80-0
DOIs	https://doi.org/10.18653/v1/2021.iwpt-1.4
Publication status	Published - 6 Aug 2021
Event	The 17th International Conference on Parsing Technologies - Bangkok, Thailand Duration: 6 Aug 2021 → 6 Aug 2021 Conference number: 17 https://iwpt21.sigparse.org/

Conference

Conference	The 17th International Conference on Parsing Technologies
Abbreviated title	IWPT 2021
Country/Territory	Thailand
City	Bangkok
Period	6/08/21 → 6/08/21
Internet address	https://iwpt21.sigparse.org/

Access to Document

10.18653/v1/2021.iwpt-1.4Licence: Creative Commons: Attribution (CC-BY)

Semi-Automatic_LI_DOA31052021_VOR_CC-BYFinal published version, 1.18 MBLicence: Creative Commons: Attribution (CC-BY)

https://aclanthology.org/2021.iwpt-1.4/Licence: Creative Commons: Attribution (CC-BY)

Cite this

Li, T., Li, S., & Steedman, M. (2021). Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer. In S. Oepen, K. Sagae, R. Tsarfaty, G. Bouma, D. Seddah, & D. Zeman (Eds.), Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021) (pp. 38-49). Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/2021.iwpt-1.4

Li, Tianyi ; Li, Sujian ; Steedman, Mark. / Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer. Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021). editor / Stephan Oepen ; Kenji Sagae ; Reut Tsarfaty ; Gosse Bouma ; Djamé Seddah ; Daniel Zeman. Stroudsburg, PA, United States : Association for Computational Linguistics (ACL), 2021. pp. 38-49

@inproceedings{a2bec9d3f44a4f819ea7d5e505118a40,

title = "Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer",

abstract = "Strong and affordable in-domain data is a desirable asset when transferring trained semantic parsers to novel domains. As previous methods for semi-automatically constructing such data cannot handle the complexity of realistic SQL queries, we propose to construct SQL queries via context-dependent sampling, and introduce the concept of topic. Along with our SQL query construction method, we propose a novel pipeline of semi-automatic Text-to-SQL dataset construction that covers the broad space of SQL queries. We show that the created dataset is comparable with expert annotation along multiple dimensions, and is capable of improving domain transfer performance for SOTA semantic parsers.",

author = "Tianyi Li and Sujian Li and Mark Steedman",

year = "2021",

month = aug,

day = "6",

doi = "10.18653/v1/2021.iwpt-1.4",

language = "English",

pages = "38--49",

editor = "Stephan Oepen and Kenji Sagae and Reut Tsarfaty and Gosse Bouma and Djam{\'e} Seddah and Daniel Zeman",

booktitle = "Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021)",

publisher = "Association for Computational Linguistics (ACL)",

note = "The 17th International Conference on Parsing Technologies, IWPT 2021 ; Conference date: 06-08-2021 Through 06-08-2021",

url = "https://iwpt21.sigparse.org/",

}

Li, T, Li, S & Steedman, M 2021, Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer. in S Oepen, K Sagae, R Tsarfaty, G Bouma, D Seddah & D Zeman (eds), Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021). Association for Computational Linguistics (ACL), Stroudsburg, PA, United States, pp. 38-49, The 17th International Conference on Parsing Technologies, Bangkok, Thailand, 6/08/21. https://doi.org/10.18653/v1/2021.iwpt-1.4

Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer. / Li, Tianyi; Li, Sujian; Steedman, Mark.

Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021). ed. / Stephan Oepen; Kenji Sagae; Reut Tsarfaty; Gosse Bouma; Djamé Seddah; Daniel Zeman. Stroudsburg, PA, United States : Association for Computational Linguistics (ACL), 2021. p. 38-49.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer

AU - Li, Tianyi

AU - Li, Sujian

AU - Steedman, Mark

N1 - Conference code: 17

PY - 2021/8/6

Y1 - 2021/8/6

N2 - Strong and affordable in-domain data is a desirable asset when transferring trained semantic parsers to novel domains. As previous methods for semi-automatically constructing such data cannot handle the complexity of realistic SQL queries, we propose to construct SQL queries via context-dependent sampling, and introduce the concept of topic. Along with our SQL query construction method, we propose a novel pipeline of semi-automatic Text-to-SQL dataset construction that covers the broad space of SQL queries. We show that the created dataset is comparable with expert annotation along multiple dimensions, and is capable of improving domain transfer performance for SOTA semantic parsers.

AB - Strong and affordable in-domain data is a desirable asset when transferring trained semantic parsers to novel domains. As previous methods for semi-automatically constructing such data cannot handle the complexity of realistic SQL queries, we propose to construct SQL queries via context-dependent sampling, and introduce the concept of topic. Along with our SQL query construction method, we propose a novel pipeline of semi-automatic Text-to-SQL dataset construction that covers the broad space of SQL queries. We show that the created dataset is comparable with expert annotation along multiple dimensions, and is capable of improving domain transfer performance for SOTA semantic parsers.

U2 - 10.18653/v1/2021.iwpt-1.4

DO - 10.18653/v1/2021.iwpt-1.4

M3 - Conference contribution

SP - 38

EP - 49

BT - Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021)

A2 - Oepen, Stephan

A2 - Sagae, Kenji

A2 - Tsarfaty, Reut

A2 - Bouma, Gosse

A2 - Seddah, Djamé

A2 - Zeman, Daniel

PB - Association for Computational Linguistics (ACL)

CY - Stroudsburg, PA, United States

T2 - The 17th International Conference on Parsing Technologies

Y2 - 6 August 2021 through 6 August 2021

ER -

Li T, Li S, Steedman M. Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer. In Oepen S, Sagae K, Tsarfaty R, Bouma G, Seddah D, Zeman D, editors, Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021). Stroudsburg, PA, United States: Association for Computational Linguistics (ACL). 2021. p. 38-49 doi: 10.18653/v1/2021.iwpt-1.4

Semi-Automatic Construction of Text-to-SQL Dataset for Domain Transfer

Abstract

Conference

Access to Document

Fingerprint

Cite this