Encoding Database Schemas with Relation-Aware Self-Attention for Text-to-SQL Parsers

Richard Shin

arxiv: 1906.11790 · v1 · pith:ZSORUYFZnew · submitted 2019-06-27 · 💻 cs.LG · cs.CL· stat.ML

Encoding Database Schemas with Relation-Aware Self-Attention for Text-to-SQL Parsers

Richard Shin This is my paper

Pith reviewed 2026-05-25 14:32 UTC · model grok-4.3

classification 💻 cs.LG cs.CLstat.ML

keywords text-to-SQL parsingrelation-aware self-attentiondatabase schema encodingSpider datasetneural encoder-decodernatural language interfaces to databases

0 comments

The pith

Relation-aware self-attention lets the encoder reason about table and column relations when turning questions into SQL.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces relation-aware self-attention inside the encoder of a text-to-SQL model. This change lets the model use information about how tables and columns connect when it processes a natural language question. The method is evaluated on the Spider dataset of complex questions over unseen database schemas. If the approach works, models can handle new database structures without needing the relations to be discovered purely from question text.

Core claim

Relation-aware self-attention within the encoder enables reasoning about how the tables and columns in the provided schema relate to each other and uses this information when interpreting the question, reaching 42.94 percent exact match accuracy on Spider versus the 18.96 percent in prior published work.

What carries the argument

relation-aware self-attention, which augments standard self-attention to incorporate explicit relational information between schema elements during encoding.

If this is right

The encoder can directly exploit foreign-key links and table connections instead of inferring them only from question wording.
Exact-match performance rises on questions that span multiple tables in a schema.
Generalization improves to database schemas and domains absent from the training set.
No extra hand-crafted schema features are required beyond the relation labels supplied to the attention layer.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same attention modification could be applied to other structured-input tasks such as semantic parsing over knowledge bases.
Collecting training data with deliberately varied relation patterns might increase the method's robustness beyond what Spider provides.
If the learned relations prove transferable, the approach could reduce the amount of schema-specific fine-tuning needed for new databases.

Load-bearing premise

The training examples supply enough different schema relations for the attention parameters to learn patterns that transfer to new schemas.

What would settle it

Accuracy on a held-out set of schemas whose relation types or combinations do not appear in the training distribution, where the reported gain over baseline encoders vanishes.

Figures

Figures reproduced from arXiv: 1906.11790 by Richard Shin.

**Figure 2.** Figure 2: An illustration of an example schema as a graph. We do not depict all edges and label types [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: Overview of the stages of our approach. Formally, we perform the following: (c fwd i,0 , c rev i,0 ), · · · ,(c fwd i,|ci| , c rev i,|ci| ) = BiLSTMColumn(c type i , ci,1, · · · , ci,|ci|); c init i = Concat(c fwd i,|ci| , c rev i,0 ) (t fwd i,1 , t rev i,1 ), · · · ,(t fwd i,|ti| , t rev i,|ti| ) = BiLSTMTable(ti,1, · · · , ti,|ti|); t init i = Concat(t fwd i,|ci| , t rev i,1 ) (q fwd 1 , q rev 1 ), · · ·… view at source ↗

read the original abstract

When translating natural language questions into SQL queries to answer questions from a database, we would like our methods to generalize to domains and database schemas outside of the training set. To handle complex questions and database schemas with a neural encoder-decoder paradigm, it is critical to properly encode the schema as part of the input with the question. In this paper, we use relation-aware self-attention within the encoder so that it can reason about how the tables and columns in the provided schema relate to each other and use this information in interpreting the question. We achieve significant gains on the recently-released Spider dataset with 42.94% exact match accuracy, compared to the 18.96% reported in published work.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Relation-aware self-attention on schemas lifts Spider exact match from 19% to 43%, but the abstract supplies almost no experimental controls or ablations.

read the letter

The main thing here is that encoding the schema with relation-aware self-attention produces a large reported gain on Spider. The abstract presents this as the key change that lets the encoder model table and column relations when interpreting the question, and the number moves from the prior 18.96% to 42.94% exact match. That is the concrete result to evaluate first. The technical step itself is straightforward: the attention mechanism is extended so that it can attend over explicit relations among schema elements rather than treating them as an unstructured bag. This targets a known pain point in text-to-SQL where complex schemas make it hard for the model to track foreign keys and join paths. The paper earns credit for naming that bottleneck clearly and for testing the idea on the recent Spider benchmark, which stresses cross-domain generalization. The soft spots sit in the evaluation. The abstract states the accuracy figure but gives no architecture diagram, no description of the decoder, no list of baselines beyond the single published number, no mention of data splits or multiple runs, and no ablation that isolates the relation-aware component. Without those, it is impossible to tell whether the jump comes from the new attention or from other unstated changes in training or model size. The stress-test worry about schema-relation overlap also stands on the given text. Spider guarantees unseen databases at test time, yet if the foreign-key graphs and column co-occurrence patterns in the test schemas resemble those in training, the model could succeed by memorizing recurring motifs rather than learning general relational reasoning. The paper would need to address that directly, either with schema-similarity controls or by showing that the learned attention weights transfer in a meaningful way. This work is for people already building neural text-to-SQL systems who want a practical tweak to schema encoding. A reader in that subfield would get value from trying the attention modification even if the current write-up is thin. It deserves peer review because the reported improvement is large enough to be worth checking, and the underlying idea is simple enough that referees can focus on whether the experiments actually support the generalization claim.

Referee Report

2 major / 1 minor

Summary. The paper proposes using relation-aware self-attention within the encoder of a text-to-SQL model so that the encoder can reason about relations among tables and columns in the input schema when interpreting the natural language question. It reports achieving 42.94% exact match accuracy on the Spider dataset, a substantial improvement over the 18.96% in prior published work.

Significance. If the experimental results hold, the work would establish that explicitly encoding schema relations via attention yields meaningful gains in cross-domain semantic parsing, directly addressing the challenge of generalizing to unseen database schemas.

major comments (2)

[Abstract] Abstract: The central accuracy claim of 42.94% exact match is stated without any description of experimental details, baselines, data splits, error bars, or training protocol, so the reported gains cannot be verified from the provided text.
[Experiments] Experiments section: The Spider test split only guarantees unseen databases, not unseen relation distributions (foreign-key graphs, join patterns, column-type co-occurrences). This leaves open the possibility that the model fits recurring training-schema motifs rather than learning transferable relational reasoning, which is the load-bearing premise of the generalization claim.

minor comments (1)

The abstract would benefit from a one-sentence outline of the model architecture or the precise form of the relation-aware attention to give readers immediate context for the accuracy number.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below, agreeing where the critique is valid and outlining specific revisions to strengthen the paper.

read point-by-point responses

Referee: [Abstract] Abstract: The central accuracy claim of 42.94% exact match is stated without any description of experimental details, baselines, data splits, error bars, or training protocol, so the reported gains cannot be verified from the provided text.

Authors: We agree that the abstract should provide sufficient context for the key result. In the revised version we will expand the abstract to briefly state that results are reported on the Spider development set using exact match accuracy, with comparison to the 18.96% baseline from prior published work, and that full experimental details (including data splits, training protocol, and model configuration) appear in the Experiments section. If error bars were computed they will be referenced; otherwise the abstract will note that the primary metric is exact match. revision: yes
Referee: [Experiments] Experiments section: The Spider test split only guarantees unseen databases, not unseen relation distributions (foreign-key graphs, join patterns, column-type co-occurrences). This leaves open the possibility that the model fits recurring training-schema motifs rather than learning transferable relational reasoning, which is the load-bearing premise of the generalization claim.

Authors: We acknowledge this limitation of the Spider benchmark: while databases are unseen, certain relational patterns may recur across train and test schemas. The 24-point absolute gain from adding relation-aware self-attention nevertheless provides evidence that the encoder benefits from explicit modeling of schema relations in a manner that improves generalization over prior approaches. We will add a short discussion paragraph noting this benchmark limitation and suggesting that future datasets with more controlled schema diversity would further isolate the contribution of relational reasoning. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical result on external benchmark with no definitional reduction

full rationale

The paper presents a neural encoder-decoder model that augments self-attention with schema relation encodings and reports 42.94% exact-match accuracy on the Spider test split. No equations, fitted parameters, or uniqueness theorems are shown that would make the accuracy score equivalent to its training inputs by construction. The benchmark split, evaluation metric, and baseline comparison (18.96%) are externally defined and independent of the model's learned weights. No self-citation chains or ansatzes are invoked to justify core claims. The derivation is therefore self-contained as a standard empirical ML contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; ledger left empty.

pith-pipeline@v0.9.0 · 5644 in / 997 out tokens · 21215 ms · 2026-05-25T14:32:06.643675+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages · 6 internal anchors

[1]

Salesforce, March 2019

A large annotated semantic parsing corpus for developing natural language interfaces.: Salesforce/WikiSQL. Salesforce, March 2019. URL https://github.com/salesforce/ WikiSQL

work page 2019
[2]

Natural Language Interfaces to Databases - An Introduction

I. Androutsopoulos, G. D. Ritchie, and P. Thanisch. Natural Language Interfaces to Databases - An Introduction. arXiv:cmp-lg/9503016, March 1995. URL http://arxiv.org/abs/cmp- lg/9503016

work page internal anchor Pith review Pith/arXiv arXiv 1995
[3]

Coarse-to-Fine Decoding for Neural Semantic Parsing

Li Dong and Mirella Lapata. Coarse-to-Fine Decoding for Neural Semantic Parsing. arXiv:1805.04793 [cs], May 2018. URL http://arxiv.org/abs/1805.04793

work page internal anchor Pith review Pith/arXiv arXiv 2018
[4]

Kummerfeld, Li Zhang, Karthik Ramanathan, Sesh Sadasivam, Rui Zhang, and Dragomir Radev

Catherine Finegan-Dollak, Jonathan K. Kummerfeld, Li Zhang, Karthik Ramanathan, Sesh Sadasivam, Rui Zhang, and Dragomir Radev. Improving Text-to-SQL Evaluation Methodology. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 351–360. Association for Computational Linguistics, 2018. URL ...

work page 2018
[5]

A Theoretically Grounded Application of Dropout in Re- current Neural Networks

Yarin Gal and Zoubin Ghahramani. A Theoretically Grounded Application of Dropout in Re- current Neural Networks. In D. D. Lee, M. Sugiyama, U. V . Luxburg, I. Guyon, and R. Garnett, editors, Advances in Neural Information Processing Systems 29 , pages 1019–1027. Curran Associates, Inc., 2016. URL http://papers.nips.cc/paper/6241-a-theoretically- grounded-...

work page 2016
[6]

Learning a neural semantic parser from user feedback

Srinivasan Iyer, Ioannis Konstas, Alvin Cheung, Jayant Krishnamurthy, and Luke Zettlemoyer. Learning a neural semantic parser from user feedback. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 963–973, 2017. URL http://www.aclweb.org/anthology/P17-1089

work page 2017
[7]

Adam: A Method for Stochastic Optimization

Diederik P. Kingma and Jimmy Ba. Adam: A Method for Stochastic Optimization. arXiv:1412.6980 [cs], December 2014. URL http://arxiv.org/abs/1412.6980

work page internal anchor Pith review Pith/arXiv arXiv 2014
[8]

Fei Li and H. V . Jagadish. Constructing an interactive natural language interface for relational databases. Proceedings of the VLDB Endowment, 8(1):73–84, September 2014. URL http: //dx.doi.org/10.14778/2735461.2735468

work page doi:10.14778/2735461.2735468 2014
[9]

Automatic differentiation in PyTorch

Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in PyTorch. October 2017. URL https://openreview.net/forum?id=BJJsrmfCZ

work page 2017
[10]

Towards a theory of natural language interfaces to databases

Ana-Maria Popescu, Oren Etzioni, , and Henry Kautz. Towards a theory of natural language interfaces to databases. In Proceedings of the 8th International Conference on Intelligent User Interfaces, pages 149–157, 2003. URL http://doi.acm.org/10.1145/604045.604070

work page doi:10.1145/604045.604070 2003
[11]

Mod- ern Natural Language Interfaces to Databases: Composing Statistical Parsing with Semantic Tractability

Ana-Maria Popescu, Alex Armanasu, Oren Etzioni, David Ko, and Alexander Yates. Mod- ern Natural Language Interfaces to Databases: Composing Statistical Parsing with Semantic Tractability. In COLING 2004: Proceedings of the 20th International Conference on Computa- tional Linguistics, 2004. URL http://aclweb.org/anthology/C04-1021

work page 2004
[12]

Self-Attention with Relative Position Representations

Peter Shaw, Jakob Uszkoreit, and Ashish Vaswani. Self-Attention with Relative Position Representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) , pages 464–468. Association for Computational Linguistics, 2018. doi: 10.18653/v1...

work page doi:10.18653/v1/n18-2074 2018
[13]

IncSQL: Training Incremental Text-to-SQL Parsers with Non-Deterministic Oracles

Tianze Shi, Kedar Tatwawadi, Kaushik Chakrabarti, Yi Mao, Oleksandr Polozov, and Weizhu Chen. IncSQL: Training Incremental Text-to-SQL Parsers with Non-Deterministic Oracles. arXiv:1809.05054 [cs], September 2018. URL http://arxiv.org/abs/1809.05054

work page internal anchor Pith review Pith/arXiv arXiv 2018
[14]

Tang and Raymond J

Lappoon R. Tang and Raymond J. Mooney. Automated construction of database interfaces: Intergrating statistical and relational learning for semantic parsing. In 2000 Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora, pages 133–141, 2000. URL http://www.aclweb.org/anthology/W00-1317. 9

work page 2000
[15]

Attention is All you Need

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is All you Need. In I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors,Advances in Neural Information Processing Systems 30, pages 5998–6008. Curran Associates, Inc., 2017....

work page 2017
[16]

SQLNet: Generating Structured Queries From Natural Language Without Reinforcement Learning

Xiaojun Xu, Chang Liu, and Dawn Song. SQLNet: Generating Structured Queries From Natural Language Without Reinforcement Learning. arXiv:1711.04436 [cs], November 2017. URL http://arxiv.org/abs/1711.04436

work page internal anchor Pith review Pith/arXiv arXiv 2017
[17]

Sqlizer: Query synthesis from natural language

Navid Yaghmazadeh, Yuepeng Wang, Isil Dillig, , and Thomas Dillig. Sqlizer: Query synthesis from natural language. In International Conference on Object-Oriented Programming, Systems, Languages, and Applications, ACM, pages 63:1–63:26, October 2017. URL http://doi.org/ 10.1145/3133887

work page doi:10.1145/3133887 2017
[18]

A Syntactic Neural Model for General-Purpose Code Generation

Pengcheng Yin and Graham Neubig. A Syntactic Neural Model for General-Purpose Code Generation. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 440–450. Association for Computational Linguistics,

work page
[19]

URL http://aclweb.org/anthology/P17-1041

doi: 10.18653/v1/P17-1041. URL http://aclweb.org/anthology/P17-1041

work page doi:10.18653/v1/p17-1041
[20]

TypeSQL: Knowledge-Based Type-Aware Neural Text-to-SQL Generation

Tao Yu, Zifan Li, Zilin Zhang, Rui Zhang, and Dragomir Radev. TypeSQL: Knowledge-Based Type-Aware Neural Text-to-SQL Generation. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 588–594. Association for Computational Linguis- ti...

work page doi:10.18653/v1/n18-2093 2018
[21]

SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-Domain Text-to-SQL Task

Tao Yu, Michihiro Yasunaga, Kai Yang, Rui Zhang, Dongxu Wang, Zifan Li, and Dragomir Radev. SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-Domain Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1653–1663. Association for Computational Linguistics, 2018. URL http: //aclweb.org/ant...

work page 2018
[22]

Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task

Tao Yu, Rui Zhang, Kai Yang, Michihiro Yasunaga, Dongxu Wang, Zifan Li, James Ma, Irene Li, Qingning Yao, Shanelle Roman, Zilin Zhang, and Dragomir Radev. Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing,...

work page 2018
[23]

Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task

Tao Yu, Rui Zhang, Kai Yang, Michihiro Yasunaga, Dongxu Wang, Zifan Li, James Ma, Irene Li, Qingning Yao, Shanelle Roman, Zilin Zhang, and Dragomir Radev. Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing,...

work page 2018
[24]

Zelle and Raymond J

John M. Zelle and Raymond J. Mooney. Learning to parse database queries using inductive logic programming. In Proceedings of the Thirteenth National Conference on Artiﬁcial Intelligence - Volume 2, pages 1050–1055, 1996. URL http://dl.acm.org/citation.cfm?id=1864519. 1864543

work page 1996
[25]

Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning

Victor Zhong, Caiming Xiong, and Richard Socher. Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning. arXiv:1709.00103 [cs], August 2017. URL http://arxiv.org/abs/1709.00103. 10

work page internal anchor Pith review Pith/arXiv arXiv 2017

[1] [1]

Salesforce, March 2019

A large annotated semantic parsing corpus for developing natural language interfaces.: Salesforce/WikiSQL. Salesforce, March 2019. URL https://github.com/salesforce/ WikiSQL

work page 2019

[2] [2]

Natural Language Interfaces to Databases - An Introduction

I. Androutsopoulos, G. D. Ritchie, and P. Thanisch. Natural Language Interfaces to Databases - An Introduction. arXiv:cmp-lg/9503016, March 1995. URL http://arxiv.org/abs/cmp- lg/9503016

work page internal anchor Pith review Pith/arXiv arXiv 1995

[3] [3]

Coarse-to-Fine Decoding for Neural Semantic Parsing

Li Dong and Mirella Lapata. Coarse-to-Fine Decoding for Neural Semantic Parsing. arXiv:1805.04793 [cs], May 2018. URL http://arxiv.org/abs/1805.04793

work page internal anchor Pith review Pith/arXiv arXiv 2018

[4] [4]

Kummerfeld, Li Zhang, Karthik Ramanathan, Sesh Sadasivam, Rui Zhang, and Dragomir Radev

Catherine Finegan-Dollak, Jonathan K. Kummerfeld, Li Zhang, Karthik Ramanathan, Sesh Sadasivam, Rui Zhang, and Dragomir Radev. Improving Text-to-SQL Evaluation Methodology. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 351–360. Association for Computational Linguistics, 2018. URL ...

work page 2018

[5] [5]

A Theoretically Grounded Application of Dropout in Re- current Neural Networks

Yarin Gal and Zoubin Ghahramani. A Theoretically Grounded Application of Dropout in Re- current Neural Networks. In D. D. Lee, M. Sugiyama, U. V . Luxburg, I. Guyon, and R. Garnett, editors, Advances in Neural Information Processing Systems 29 , pages 1019–1027. Curran Associates, Inc., 2016. URL http://papers.nips.cc/paper/6241-a-theoretically- grounded-...

work page 2016

[6] [6]

Learning a neural semantic parser from user feedback

Srinivasan Iyer, Ioannis Konstas, Alvin Cheung, Jayant Krishnamurthy, and Luke Zettlemoyer. Learning a neural semantic parser from user feedback. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 963–973, 2017. URL http://www.aclweb.org/anthology/P17-1089

work page 2017

[7] [7]

Adam: A Method for Stochastic Optimization

Diederik P. Kingma and Jimmy Ba. Adam: A Method for Stochastic Optimization. arXiv:1412.6980 [cs], December 2014. URL http://arxiv.org/abs/1412.6980

work page internal anchor Pith review Pith/arXiv arXiv 2014

[8] [8]

Fei Li and H. V . Jagadish. Constructing an interactive natural language interface for relational databases. Proceedings of the VLDB Endowment, 8(1):73–84, September 2014. URL http: //dx.doi.org/10.14778/2735461.2735468

work page doi:10.14778/2735461.2735468 2014

[9] [9]

Automatic differentiation in PyTorch

Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in PyTorch. October 2017. URL https://openreview.net/forum?id=BJJsrmfCZ

work page 2017

[10] [10]

Towards a theory of natural language interfaces to databases

Ana-Maria Popescu, Oren Etzioni, , and Henry Kautz. Towards a theory of natural language interfaces to databases. In Proceedings of the 8th International Conference on Intelligent User Interfaces, pages 149–157, 2003. URL http://doi.acm.org/10.1145/604045.604070

work page doi:10.1145/604045.604070 2003

[11] [11]

Mod- ern Natural Language Interfaces to Databases: Composing Statistical Parsing with Semantic Tractability

Ana-Maria Popescu, Alex Armanasu, Oren Etzioni, David Ko, and Alexander Yates. Mod- ern Natural Language Interfaces to Databases: Composing Statistical Parsing with Semantic Tractability. In COLING 2004: Proceedings of the 20th International Conference on Computa- tional Linguistics, 2004. URL http://aclweb.org/anthology/C04-1021

work page 2004

[12] [12]

Self-Attention with Relative Position Representations

Peter Shaw, Jakob Uszkoreit, and Ashish Vaswani. Self-Attention with Relative Position Representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) , pages 464–468. Association for Computational Linguistics, 2018. doi: 10.18653/v1...

work page doi:10.18653/v1/n18-2074 2018

[13] [13]

IncSQL: Training Incremental Text-to-SQL Parsers with Non-Deterministic Oracles

Tianze Shi, Kedar Tatwawadi, Kaushik Chakrabarti, Yi Mao, Oleksandr Polozov, and Weizhu Chen. IncSQL: Training Incremental Text-to-SQL Parsers with Non-Deterministic Oracles. arXiv:1809.05054 [cs], September 2018. URL http://arxiv.org/abs/1809.05054

work page internal anchor Pith review Pith/arXiv arXiv 2018

[14] [14]

Tang and Raymond J

Lappoon R. Tang and Raymond J. Mooney. Automated construction of database interfaces: Intergrating statistical and relational learning for semantic parsing. In 2000 Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora, pages 133–141, 2000. URL http://www.aclweb.org/anthology/W00-1317. 9

work page 2000

[15] [15]

Attention is All you Need

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is All you Need. In I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors,Advances in Neural Information Processing Systems 30, pages 5998–6008. Curran Associates, Inc., 2017....

work page 2017

[16] [16]

SQLNet: Generating Structured Queries From Natural Language Without Reinforcement Learning

Xiaojun Xu, Chang Liu, and Dawn Song. SQLNet: Generating Structured Queries From Natural Language Without Reinforcement Learning. arXiv:1711.04436 [cs], November 2017. URL http://arxiv.org/abs/1711.04436

work page internal anchor Pith review Pith/arXiv arXiv 2017

[17] [17]

Sqlizer: Query synthesis from natural language

Navid Yaghmazadeh, Yuepeng Wang, Isil Dillig, , and Thomas Dillig. Sqlizer: Query synthesis from natural language. In International Conference on Object-Oriented Programming, Systems, Languages, and Applications, ACM, pages 63:1–63:26, October 2017. URL http://doi.org/ 10.1145/3133887

work page doi:10.1145/3133887 2017

[18] [18]

A Syntactic Neural Model for General-Purpose Code Generation

Pengcheng Yin and Graham Neubig. A Syntactic Neural Model for General-Purpose Code Generation. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 440–450. Association for Computational Linguistics,

work page

[19] [19]

URL http://aclweb.org/anthology/P17-1041

doi: 10.18653/v1/P17-1041. URL http://aclweb.org/anthology/P17-1041

work page doi:10.18653/v1/p17-1041

[20] [20]

TypeSQL: Knowledge-Based Type-Aware Neural Text-to-SQL Generation

Tao Yu, Zifan Li, Zilin Zhang, Rui Zhang, and Dragomir Radev. TypeSQL: Knowledge-Based Type-Aware Neural Text-to-SQL Generation. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 588–594. Association for Computational Linguis- ti...

work page doi:10.18653/v1/n18-2093 2018

[21] [21]

SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-Domain Text-to-SQL Task

Tao Yu, Michihiro Yasunaga, Kai Yang, Rui Zhang, Dongxu Wang, Zifan Li, and Dragomir Radev. SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-Domain Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1653–1663. Association for Computational Linguistics, 2018. URL http: //aclweb.org/ant...

work page 2018

[22] [22]

Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task

Tao Yu, Rui Zhang, Kai Yang, Michihiro Yasunaga, Dongxu Wang, Zifan Li, James Ma, Irene Li, Qingning Yao, Shanelle Roman, Zilin Zhang, and Dragomir Radev. Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing,...

work page 2018

[23] [23]

Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task

Tao Yu, Rui Zhang, Kai Yang, Michihiro Yasunaga, Dongxu Wang, Zifan Li, James Ma, Irene Li, Qingning Yao, Shanelle Roman, Zilin Zhang, and Dragomir Radev. Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing,...

work page 2018

[24] [24]

Zelle and Raymond J

John M. Zelle and Raymond J. Mooney. Learning to parse database queries using inductive logic programming. In Proceedings of the Thirteenth National Conference on Artiﬁcial Intelligence - Volume 2, pages 1050–1055, 1996. URL http://dl.acm.org/citation.cfm?id=1864519. 1864543

work page 1996

[25] [25]

Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning

Victor Zhong, Caiming Xiong, and Richard Socher. Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning. arXiv:1709.00103 [cs], August 2017. URL http://arxiv.org/abs/1709.00103. 10

work page internal anchor Pith review Pith/arXiv arXiv 2017