Proj AutoWriter Paper Reading: SEQ2SQL: GENERATING STRUCTURED QUERIES FROM NATURAL LANGUAGE USING REINFORCEMENT LEARNING(Proj AutoWriter Paper Reading: SEQ2SQL: GENERATING STRUCTURED QUERIES FROM NATURAL LANGUAGE USING REINFORCEMENT LEARNING)

Abstract

  • 介绍Relational DB
    本文Seq2SQL
    方法: policy-based reinforcement learning
    Task:…
  • 使用in-the-loop query执行来获取rewards
  • 利用SQL的结构来剪枝空间
    本文: 数据集: WikiSQL
    规模80654个人工标注的问题,24241个表
    实验:
  • 比最好的semantic parser更好
  • improving execution accuracy from 35.9% to 59.4%
  • logical form accuracy from 23.4% to 48.3%.
————————

Abstract

  • 介绍Relational DB
    本文Seq2SQL
    方法: policy-based reinforcement learning
    Task:…
  • 使用in-the-loop query执行来获取rewards
  • Pruning space using SQL structure
    Article: dataset: wikisql
    Scale 80654 manually marked problems and 24241 tables
    experiment:
  • 比最好的semantic parser更好
  • improving execution accuracy from 35.9% to 59.4%
  • logical form accuracy from 23.4% to 48.3%.