LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 2541–2555 August 1–6, 2021. ©2021 Association for Computational Linguistics 2541 LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations Ruisheng Cao1, Lu Chen1,2∗, Zhi Chen1, Yanbin Zhao1, Su Zhu3 and Kai Yu1,2∗ 1X-LANCE Lab, Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence, AI Institute, Shanghai Jiao Tong University Shanghai Jiao Tong University, Shanghai, China 2State Key Lab of Media Convergence Production Technology and Systems, Beijing, China 3AISpeech Co., Ltd., Suzhou, China {211314,chenlusz,kai.yu}@sjtu.edu.cn Abstract This work aims to tackle the challenging het- erogeneous graph encoding problem in the text-to-SQL task. Previous methods are typi- cally node-centric and merely utilize different weight matrices to parameterize edge types, which 1) ignore the rich semantics embed- ded in the topological structure of edges, and 2) fail to distinguish local and non- local relations for each node. To this end, we propose a Line Graph Enhanced Text-to- SQL (LGESQL) model to mine the underlying relational features without constructing meta- paths. By virtue of the line graph, messages propagate more efficiently through not only connections between nodes, but also the topol- ogy of directed edges. Furthermore, both lo- cal and non-local relations are integrated dis- tinctively during the graph iteration. We also design an auxiliary task called graph pruning to improve the discriminative capability of the encoder. Our framework achieves state-of-the- art results (62.8% with GLOVE, 72.0% with ELECTRA) on the cross-domain text-to-SQL benchmark Spider at the time of writing. 1 Introduction The text-to-SQL task (Zhong et al., 2017; Xu et al., 2017) aims to convert a natural language question into a SQL query, given the corresponding database schema. It has been widely studied in both aca- demic and industrial communities to build natural language interfaces to databases (NLIDB, Androut- sopoulos et al., 1995). One daunting problem is how to jointly encode the question words and database schema items (in- cluding tables and columns), as well as various relations among these heterogeneous inputs. Typ- ically, previous literature utilizes a node-centric graph neural network (GNN, Scarselli et al., 2008) ∗The corresponding authors are Lu Chen and Kai Yu. Figure 1: Two limitations if edge features are retrieved from a fixed-size embedding matrix: (a) fail to discover useful meta-paths, and (b) unable to differentiate local and non-local neighbors. to aggregate information from neighboring nodes. GNNSQL (Bogin et al., 2019a) adopts a relational graph convolution network (RGCN, Schlichtkrull et al., 2018) to take into account different edge types between schema items, such as T-HAS-C relationship 1, primary key and foreign key con- straints. However, these edge features are directly retrieved from a fixed-size parameter matrix and may suffer from the drawback: unaware of con- textualized information, especially the structural topology of edges. Meta-path is defined as a com- posite relation linking two objects, which can be used to capture multi-hop semantics. For example, in Figure 1(a), relation Q-EXACTMATCH-C and …

Content truncated to 500 words