Abstract:Recently, the Text-to-SQL task has received much attention. Many sophisticated neural models have been invented that achieve significant results. Most current work assumes that all the inputs are legal and the model should generate an SQL query for any input. However, in the real scenario, users are allowed to enter the arbitrary text that may not be answered by an SQL query. In this article, we focus on the issue–answerability classification for the Text-to-SQL system, which aims to distinguish the answerability of the question according to the given database schema. Existing methods concatenate the question and the database schema into a sentence, then fine-tune the pre-trained language model on the answerability classification task. In this way, the database schema is regarded as sequence text that may ignore the intrinsic structure relationship of the schema data, and the attention that represents the correlation between the question token and the database schema items is not well designed. To this end, we propose a relational Question-Schema graph framework that can effectively model the attention and relation between question and schema. In addition, a conditional layer normalization mechanism is employed to modulate the pre-trained language model to generate better question representation. Experiments demonstrate that the proposed framework outperforms all existing models by large margins, achieving new state of the art on the benchmark TRIAGESQL. Specifically, the model attains 88.41%, 78.24%, and 75.98% in Precision, Recall, and F1, respectively. Additionally, it outperforms the baseline by approximately 4.05% in Precision, 6.96% in Recall, and 6.01% in F1.

RGISQL: Integrating Refined Grammatical Information into Relational Graph Neural Network for Text-to-SQL Task

IGSQL: Database Schema Interaction Graph Based Neural Model for Context-Dependent Text-to-SQL Generation

IGSQL: Database Schema Interaction Graph Based Neural Model for Context-Dependent Text-to-SQL Generation

An interaction-modeling mechanism for context-dependent Text-to-SQL translation based on heterogeneous graph aggregation

RSL-SQL: Robust Schema Linking in Text-to-SQL Generation

Relation-Aware Graph Transformer for SQL-to-Text Generation

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

A Heterogeneous Graph to Abstract Syntax Tree Framework for Text-to-SQL

QURG: Question Rewriting Guided Context-Dependent Text-to-SQL Semantic Parsing

Towards Complex Text-to-SQL in Cross-Domain Database with Intermediate Representation

ShadowGNN: Graph Projection Neural Network for Text-to-SQL Parser.

Semantic Enhanced Text-to-SQL Parsing Via Iteratively Learning Schema Linking Graph

S$^2$SQL: Injecting Syntax to Question-Schema Interaction Graph Encoder for Text-to-SQL Parsers

Grammar-based Neural Text-to-SQL Generation

ER-SQL: Learning Enhanced Representation for Text-to-SQL Using Table Contents.

HIE-SQL: History Information Enhanced Network for Context-Dependent Text-to-SQL Semantic Parsing

Bravely Say I Don’t Know: Relational Question-Schema Graph for Text-to-SQL Answerability Classification

Graph Alignment for Cross-Domain Text-to-SQL

CQR-SQL: Conversational Question Reformulation Enhanced Context-Dependent Text-to-SQL Parsers

RH-SQL: Refined Schema and Hardness Prompt for Text-to-SQL

Bertrand-DR: Improving Text-to-SQL using a Discriminative Re-ranker