![](https://img-blog.csdnimg.cn/20201014180756922.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
Text-to-SQL
文章平均质量分 93
Q同学的nlp笔记
深度解读最新NLP论文,关注技术动态,分享NLP在各行各业的应用见解。
展开
-
Text2SQL学习整理(一) 综述
Text2SQL是近年来NLP领域一个比较热门的研究方向,该任务历史悠久,应用和落地性很强。该任务是在已知数据库的表名、列名其从属关系(这些统称为数据库的Schema)的前提下,将人类的自然语言问句(Question)转化为对应的数据库查询SQL语句。原创 2023-12-17 12:10:54 · 1430 阅读 · 1 评论 -
论文笔记:CQR-SQL: Conversational Question Reformulation Enhanced Context-Dependent Text-to-SQL Parsers
对于多轮的Text-to-SQL任务,现有的方法通常专注于充分利用历史上下文或以前预测的SQL来进行当前的SQL解析,而忽略了显式地理解模式和会话依赖关系,如共同引用、省略和用户焦点变化。而这些则是多轮Text-to-SQL中最复杂的挑战,下图1展示了关于这些语言现象的示例:之前的工作大都是简单的将多轮中的各个句子进行拼接(下图2a),就得到了不错的效果,但这些端到端的方法缺乏对这些上下文依赖现象的关注。原创 2023-03-24 22:46:43 · 617 阅读 · 4 评论 -
论文笔记:HIE-SQL:History Information Enhanced Network for Context-Dependent Text-to-SQL Semantic Parsing
论文笔记:HIE-SQL: History Information Enhanced Network for Context-Dependent Text-to-SQL Semantic Parsing目录论文笔记:HIE-SQL: History Information Enhanced Network for Context-Dependent Text-to-SQL Semantic Parsing导语摘要1 简介2 相关工作3 HIE-SQL3.1 预备知识3.2 Multimodal Encod原创 2022-04-24 15:59:15 · 873 阅读 · 0 评论 -
论文笔记:Pay More Attention to History: A Context Modeling Strategy for Conversational Text-to-SQL
论文笔记:Pay More Attention to History: A Context Modeling Strategy for Conversational Text-to-SQL目录论文笔记:Pay More Attention to History: A Context Modeling Strategy for Conversational Text-to-SQL导语摘要关键词1 简介2 相关工作2.1 语义解析和Text-to-SQL任务2.2 多轮对话Text-to-SQL3 问题定义4原创 2022-04-23 22:10:16 · 433 阅读 · 0 评论 -
论文笔记:SCORE: Pre-training for Context Representation in Conversational Semantic Parsing
论文笔记:SCORE: Pre-training for Context Representation in Conversational Semantic Parsing目录论文笔记:SCORE: Pre-training for Context Representation in Conversational Semantic Parsing导语导语摘要1 简介2 方法2.1 预备知识2.2 SCORE的预训练2.3 数据合成3&4 实验与分析整体结果不同预训练目标函数的差异不同轮次之间的效果原创 2022-04-23 20:10:21 · 1635 阅读 · 0 评论 -
论文笔记:An Interactive NL2SQL Approach with Reuse Strategy
论文笔记:An Interactive NL2SQL Approach with Reuse Strategy目录论文笔记:An Interactive NL2SQL Approach with Reuse Strategy导语摘要关键词1 简介2 相关工作3 方法3.1 任务定义3.2 Tree-SQL3.3 基本模型3.4 利用重用机制进行优化导语会议:DASFAA 2021地址:https://link.springer.com/chapter/10.1007/978-3-030-73197原创 2022-04-23 12:57:09 · 944 阅读 · 0 评论 -
论文笔记:Structure-Grounded Pretraining for Text-to-SQL
论文笔记:Structure-Grounded Pretraining for Text-to-SQL目录论文笔记:Structure-Grounded Pretraining for Text-to-SQL导语导语摘要1 简介2 相关工作跨数据库的Text-to-SQLText-Table数据的预训练Text-to-SQL中的结构对齐3 结构对齐的预训练(Structure-Grounded Pretraining)3.1 动机3.2 预训练的目标列对齐值对齐列-值映射3.3 通过弱监督获取预训练数据4原创 2022-04-22 09:26:40 · 734 阅读 · 1 评论 -
论文笔记:GRAPPA: GRAMMAR-AUGMENTED PRE-TRAINING FOR TABLE SEMANTIC PARSING
论文笔记:GRAPPA: GRAMMAR-AUGMENTED PRE-TRAINING FOR TABLE SEMANTIC PARSING目录论文笔记:GRAPPA: GRAMMAR-AUGMENTED PRE-TRAINING FOR TABLE SEMANTIC PARSING导语导语会议:ICLR 2021地址:https://arxiv.org/abs/2009.13845原创 2022-04-21 22:57:24 · 1798 阅读 · 1 评论 -
论文笔记:TABERT: Pretraining for Joint Understanding of Textual and Tabular Data
论文笔记:TABERT: Pretraining for Joint Understanding of Textual and Tabular Data目录论文笔记:TABERT: Pretraining for Joint Understanding of Textual and Tabular Data导语摘要1 简介2 背景2.1 Semantic Parsing over Tables2.2 Masked Language Models3 TABERT: 从表格数据中学习联合分布3.1 计算自然语原创 2022-04-20 15:44:42 · 2317 阅读 · 0 评论 -
论文笔记:S2SQL: Injecting Syntax to Question-Schema Interaction Graph Encoder for Text-to-SQL Parsers
论文笔记:S2SQL: Injecting Syntax to Question-Schema Interaction Graph Encoder for Text-to-SQL Parsers目录论文笔记:S2SQL: Injecting Syntax to Question-Schema Interaction Graph Encoder for Text-to-SQL Parsers导语摘要1 简介2 提出的方法2.1 问题定义2.2 Question-Schema Interaction Grap原创 2022-04-14 21:37:21 · 939 阅读 · 0 评论 -
论文笔记:PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models
论文笔记:PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models目录论文笔记:PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models导语摘要1 简介2 PICARD方法2.1 Lexing2.2 Parsing without Guards2.3 Par原创 2021-09-29 17:41:35 · 1147 阅读 · 0 评论 -
论文笔记:Bridging Textual and Tabular Data for Cross-Domain Text-to-SQL Semantic Parsing
论文笔记:Bridging Textual and Tabular Data for Cross-Domain Text-to-SQL Semantic Parsing目录论文笔记:Bridging Textual and Tabular Data for Cross-Domain Text-to-SQL Semantic Parsing导语摘要1 简介2 模型2.1 问题定义2.2 Question-Schema Serialization and Encoding2.3 Bridging2.4 解码2原创 2021-09-24 12:14:40 · 709 阅读 · 0 评论 -
论文笔记:Editing-Based SQL Query Generation for Cross-Domain Context-Dependent Questions
论文笔记:Editing-Based SQL Query Generation for Cross-Domain Context-Dependent Questions目录论文笔记:Editing-Based SQL Query Generation for Cross-Domain Context-Dependent Questions笔者按摘要1 简介2 上下文相关的跨领域语义解析2.1 数据集2.2 任务定义3 方法3.1 Utterance-Table Encoder3.2 Interaction原创 2021-08-25 10:53:43 · 469 阅读 · 0 评论 -
论文笔记:RAT-SQL: Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers
论文笔记:RAT-SQL: Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers目录论文笔记:RAT-SQL: Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers笔者按摘要简介相关工作Relation-Aware Self-AttentionRAT-SQL问题定义Relation-Aware Input Encoding笔者按RATSQL是A原创 2021-08-08 10:46:54 · 2383 阅读 · 3 评论 -
论文笔记:LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations
论文笔记:LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations笔者按这篇文章是截止目前(2021.8)Spider数据集榜单上的SOTA。文章从RATSQL的基础上对Encoder部分进行了多方面改进。首先将原始的异构图根据其1阶连接关系,拆分成节点图和线图两个部分,对两个互为对偶的输入各自应用关系图注意力网络(RGAT),其次加入一个Graph pruning的附加任务辅助Text-原创 2021-08-05 11:01:49 · 2236 阅读 · 4 评论