README.md

metadata

configs:
  - config_name: default
    sep: ;
    data_files:
      - split: 1M
        path: 1M_batch.csv
      - split: train
        path: train.csv
      - split: test1
        path: test1.csv
      - split: test2
        path: test2.csv
      - split: hand_crafted
        path: hand_crafted_testset.csv
task_categories:
  - text-generation
  - translation
language:
  - en
tags:
  - text-to-sql
  - question-to-sql
  - nlq-to-sql
  - SQL
  - english-to-sql
size_categories:
  - 100K<n<1M

BanQies

Dataset Summary

BanQies is a large-scale, complex, domain-specific text-to-SQL dataset generated using SelectCraft. The goal of SelectCraft is to enable the automatic generation of domain-specific text-to-SQL datasets, with BanQies serving as a use case in the financial domain.

Data Fields

• db_id: Name of the database associated with the query
• question: Natural language question to be translated into SQL
• query: Ground truth SQL query corresponding to the question

Data Splits

• train: 630.000 question–SQL query pairs
• test1: 70.000 question–SQL query pairs
• test2: 1000 question–SQL query pairs
• hand crafted test: 80 question–SQL query pairs

Databases

1. Loan_DB:

CREATE TABLE Loans \n{ \n loan_id number, \n client_id number, \n budget real, \n duration number, \n interest real, \n status varchar \n} \n CREATE TABLE Clients \n{ \n client_id number, \n first_name varchar, \n last_name varchar, \n email varchar, \n city varchar, \n year_of_birth number \n} \n CREATE TABLE Accounts \n{ \n account_id number, \n client_id number, \n balance real, \n type varchar \n} \n CREATE TABLE Deposits \n{ \n deposit_id number, \n account_id number, \n source varchar, \n amount real \n} \n

2. Transactions_DB:

CREATE TABLE Transactions \n{ \n transaction_id number, \n timestamp_id number, \n primary_contract_id number, \n client_id number, \n beneficiary_id number, \n transaction_amount real, \n is_fraudulent boolean, \n product_family_code varchar, \n amount_currency varchar \n} \n CREATE TABLE Beneficiary \n{ \n beneficiary_id number, \n bank_branch_id number, \n country_name varchar, \n country_code varchar \n} \n CREATE TABLE Source \n{ \n primary_contract_id number, \n client_id number, \n counterparty_bank_branch_id number, \n counterparty_donor_id number \n} \n CREATE TABLE Time \n{ \n timestamp_id number, \n week_number number, \n day_number number, \n hour_number number, \n day_name varchar, \n year number, \n month_number number \n} \n

Citation