NL2SQL

NL2SQL seeks to transform natural language questions to SQL, allowing individuals to run unstructured queries against databases.

Dependencies

• NLTK
• Stanford CoreNLP
• Plac
• Requests
• MySQLdb
• Faker
• NumPy

Quickstart

$ pip install -r requirements.txt
$ python nl2sql.py download
$ python nl2sql.py setup
$ python nl2sql.py run

Enter sentences to be turned into SQL:

$ python nl2sql.py run

  Type 'exit' to quit

  How can I help you?

  >: Find several students registered for Constitutional Law

usage: python nl2sql.py {download,setup,run} ...

Command-line interface for NL2SQL

subcommands:
  {download,setup,run}
    download            Download the Stanford CoreNLP tools and related
                        models. Install the needed project dependencies.
    setup               Setup the project as needed.
    run                 Start NL2SQL to parse questions

Download

To download Dependencies (may take a while):

$ pip install -r requirements.txt
$ python nl2sql.py download

Example run:

$ python nl2sql.py download

  Downloading stanford-corenlp-full-2016-10-31.zip from http://nlp.stanford.edu/software/stanford-corenlp-full-2016-10-31.zip

  [==================================================]
   
  Extracting stanford-corenlp-full-2016-10-31.zip to /Users/.../nl2sql/data

  Extracting stanford-corenlp-3.7.0-models.jar to /Users/.../nl2sql/data/stanford-corenlp-full-2016-10-31/stanford-corenlp-3.7.0-models

usage: python nl2sql.py download [-p PATH] [-f]

    Download the Stanford CoreNLP tools and related models.
    Install the needed project dependencies.

optional arguments:
  -p PATH, --path PATH  Path to download models and jar files (default:
                        ./data)
  -f, --force           Force download of all files

Setup

To setup, first have a MySQL database ready and running, then run:

$ python nl2sql.py setup

Which will prompt you to enter a database information, then prompt you to import the schema and seed the database.

Next, it will generate the SchemaGraph and DBCorpus and train all Classifiers and store them via pickle for later use.

Example run:

$ python nl2sql.py setup

  Setting up NL2SQL.

  Downloading WordNet corpora.

  [nltk_data] Downloading package wordnet to /...
  [nltk_data]   Package wordnet is already up-to-date!

  Configuring Database:

  Enter hostname: LOCALHOST
  Enter MySQL user: ROOT
  Enter MySQL password:
  Enter database name: DATABASE_NAME
  Database does not exist. Create it? [y/n]: y
  Do you want to import the database schema? [y/n]: y
  Do you want to seed the database? [y/n]: y

  Database configured.

  Constructing Database Graph.
  
  Database Graph constructed.
  
  Creating Database Corpus.

  Database Corpus created.

  Training database classifier.

  ==> Training (20 iterations)

      Iteration    Log Likelihood    Accuracy
      ---------------------------------------
             1          -2.70805        0.010
             2          -0.78485        0.974
             ...
            19          -0.10204        1.000
         Final          -0.09748        1.000

  Database classifier trained.

  Training SQL grammar classifier.

  SQL grammar classifier trained.

  Set up complete.

usage: python nl2sql.py setup [-f]

    Setup the project as needed.

optional arguments:
  -f, --force  Force setup to rerun

Run

To start the system, run the two previous commands, then run:

$ python nl2sql.py setup

You will be prompted to enter a statement which will be turned into SQL and executed.

Example Run:

$ python nl2sql.py run

  Type 'exit' to quit

  How can I help you?

  >: How many sections are located in Rhode Island in Spring 2017?

  +------------+
  | COUNT(*)   |
  +------------+
  | 15         |
  +------------+

   What else would you like to know?

   >:

usage: python nl2sql.py run [-d]

    Start NL2SQL to parse questions

optional arguments:
  -d, --debug  Print out debug statements