Research on Question Understanding Method Combining Database Structure and Content

doi:10.19678/j.issn.1000-3428.0057250

Computer Engineering ›› 2021, Vol. 47 ›› Issue (3): 71-76,82. doi: 10.19678/j.issn.1000-3428.0057250

• Artificial Intelligence and Pattern Recognition • Previous Articles Next Articles

Research on Question Understanding Method Combining Database Structure and Content

YUAN Zhixiang, REN Dongdong, HONG Xudong, SUN Guohua

School of Computer Science and Technology, Anhui University of Technology, Maanshan, Anhui 243000, China

Received:2020-01-17 Revised:2020-03-02 Published:2020-03-09

结合数据库结构及内容的问句理解方法研究

袁志祥, 任冬冬, 洪旭东, 孙国华

安徽工业大学计算机科学与技术学院, 安徽马鞍山 243000

作者简介:袁志祥(1973-),男,副教授、硕士,主研方向为机器学习、工业互联网、信息系统、数据分析;任冬冬,硕士研究生;洪旭东,博士;孙国华,硕士。
基金资助:
国家重点研发计划（2016YFF020440508）。

Abstract

Abstract: Question understanding is an important basis for the model to transform natural language questions into SQL.At present,most of the models using deep learning generate database queries through only the database structure and do not fully understand the question in combination with the database contents.To address this problem,based on the SQLova model,this paper proposes a question understanding method combining database structure and content.This method uses the attention mechanisms for table structure and table content to obtain more accurate semantic expressions of the question.The sub-class classification task is used to fill the SQL sketch to complete the SQL query.The model is tested on a Chinese data set released on the first Chinese NL2SQL Challenge of Alibaba Cloud.Test results show that the proposed question understanding method combining database structure and content achieves an accuracy rate of 78%, which is 1.8% higher than that of the model without table contents.When the method is tested on the WikiSQL dataset, its accuracy rate is 1.4% higher than that of SQLova.The experiment results show that the proposed method can effectively improve the accuracy of the generated SQL queries.

Key words: table structure, table content, question understanding, slot filling, SQL query

摘要： 问句理解是模型将自然语言问句转换成SQL的重要基础。目前多数利用深度学习的模型仅是通过数据库结构，未结合数据库内容充分理解问句生成SQL查询。在SQLova模型的基础上，提出一种基于表结构和内容的问句理解方法。利用表结构和表内容关注机制获得问句更准确的语义表达式，通过子类分类任务填充SQL草图完成SQL查询。在阿里云首届中文NL2SQL挑战赛发布的中文数据集上进行测试，结果表明，结合数据库结构与内容的问句理解方法取得78%的准确率，比不结合表内容的模型高出1.8%，在WikiSQL数据集上比SQLova准确率高出1.4%，可以有效提高生成SQL查询的准确率。

关键词: 表结构, 表内容, 问句理解, 槽填充, SQL查询

CLC Number:

TP18

YUAN Zhixiang, REN Dongdong, HONG Xudong, SUN Guohua. Research on Question Understanding Method Combining Database Structure and Content[J]. Computer Engineering, 2021, 47(3): 71-76,82.

袁志祥, 任冬冬, 洪旭东, 孙国华. 结合数据库结构及内容的问句理解方法研究[J]. 计算机工程, 2021, 47(3): 71-76,82.

/ / Recommend / Download Citations

URL: http://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0057250

http://www.ecice06.com/EN/Y2021/V47/I3/71

Figures/Tables 6

References

[1] ZHONG V,XIONG C,SOCHER R.Seq2SQL:generating structured queries from natural language using reinforcement learning[EB/OL].[2019-12-10].https://arxiv.org/pdf/1709.00103.pdf.
[2] GIORDANI A,MOSCHITTI A.Automatic generation and reranking of SQL-derived answers to NL questions[C]//Proceedings of International Workshop on Eternal Systems.Berlin,Germany:Springer,2012:59-76.
[3] DE MARNEFFE M C,MACCARTNEY B,MANNING C D.Generating typed dependency parses from phrase structure parses[C]//Proceedings of IEEE International Conference on Language Resources and Evaluation.Washington D.C.,USA:IEEE Press,2006:449-454.
[4] SHEN L,JOSHI A K.An SVM-based voting algorithm with application to parse reranking[C]//Proceedings of the 7th Conference on Natural Language Learning at HLT-NAACL.Washington D.C.,USA:IEEE Press,2003:9-16.
[5] POON H.Grounded unsupervised semantic parsing[C]//Proceedings of the 51st ACM Annual Meeting of the Association for Computational Linguistics.New York,USA:ACM Press,2013:933-943.
[6] LI F,JAGADISH H V.Constructing an interactive natural language interface for relational databases[J].Proceedings of the VLDB Endowment,2014,8(1):73-84.
[7] ZHOU Junsheng,QU Weiguang,XU Juhong,et al.Research on the implementation of Chinese GIS natural language interface based on semantic analysis[J].Journal of Chinese Information Processing,2014,28(6):62-69.(in Chinese)周俊生,曲维光,许菊红,等.基于语义解析的中文GIS自然语言接口实现研究[J].中文信息学报,2014,28(6):62-69.
[8] BAHDANAU D,CHO K.Neural machine translation by jointly learning to align and translate[EB/OL].[2019-12-10].https://arxiv.org/pdf/1409.0473.pdf.
[9] DONG L,LAPATA M.Language to logical form with neural attention[EB/OL].[2019-12-10].https://arxiv.org/pdf/1601.01280.pdf.
[10] XU Xiaojun,LIU Chang,SONG Dawen.SQLNET:generating structured queries from natural language without reinforcement learning[EB/OL].[2019-12-10].https://arxiv.org/pdf/1711.04436.pdf.
[11] YU Tao,LI Zofan,ZHANG Zilin,et al.Typesql:knowledge-based type-aware neural text-to-SOL generation[EB/OL].[2019-12-10].https://arxiv.org/pdf/1804.09769.pdf.
[12] SHI T,TATEAEADI K,CHAKRABARTI K,et al.IncSQL:training incremental text-to-SQL parsers with non-deterministic oracles[EB/OL].[2019-12-10].https://arxiv.org/abs/1809.05054.
[13] WANG C,TATWAWADI K,BROCKSCHMIDT M,et al.Robust text-to-SOL generation with execution-guided decoding[EB/OL].[2019-12-10].https://arxiv.org/pdf/1807.03100.pdf.
[14] HWANG W,YIM J,PARK S,et al.A comprehensive exploration on WikiSQL with table-aware word contextualization[EB/OL].[2019-12-10].https://arxiv.org/pdf/1902.01069.pdf.
[15] DEVLIN J,CHANG M W,LEE K,et al.Bert:pre-training of deep bidirectional transformers for language understanding[C]//Proceedings of 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.Minneapolis,USA:Association for Computational Linguistics 2019:4171-4186.
[16] PENNINGTON J,SOCHER R,MANNING C.GloVe:global vectors for word representation[C]//Proceedings of IEEE Conference on Empirical Methods in Natural Language Processing.Washington D.C.,USA:IEEE Press,2014:1532-1543.
[17] HE P,MAO Y,CHAKRABARTI K,et al.X-SQL:reinforce schema representation with context[EB/OL].[2019-12-10].https://arxiv.org/abs/1908.08113v1.
[18] LIU Xiaodong,HE Pengcheng,CHEN Weizhu,et al.Multi-task deep neural networks for natural language understanding[EB/OL].[2019-12-10].https://arxiv.org/pdf/1901.11504.pdf.
[19] PETERS M E,NEUMANN M,IYYER M,et al.Deep contextualized word representations[EB/OL].[2019-12-10].https://arxiv.org/pdf/1802.05365.pdf.
[20] WU Xukang,YANG Xuguang,CHEN Yuanyuan,et al.Topic combined word vector model[J].Computer Engineering,2018,44(2):233-237,270.(in Chinese)吴旭康,杨旭光,陈园园,等.主题联合词向量模型[J].计算机工程,2018,44(2):233-237,270.
[21] DONG L,LAPATA M.Coarse-to-fine decoding for neural semantic parsing[EB/OL].[2019-12-10].https://arxiv.org/pdf/1805.04793.pdf.

Please choose a citation manager

Content to export