/parsetron

A natural language semantic parser

Primary LanguagePythonApache License 2.0Apache-2.0

Parsetron - A natural language semantic parser

https://travis-ci.org/Kitt-AI/parsetron.svg?branch=develop https://coveralls.io/repos/Kitt-AI/parsetron/badge.svg?branch=develop Documentation Status

A natural language semantic parser

Installation

pip install parsetron

Parsetron is tested under Python 2.7 and Pypy. It doesn't support Python 3 yet.

Documentation and Tutorial

http://parsetron.kitt.ai

Quick Start

The following is a grammar that parses natural language instruction on lights:

from parsetron import Set, Regex, Optional, OneOrMore, Grammar, RobustParser

class LightGrammar(Grammar):

    action = Set(['change', 'flash', 'set', 'blink'])
    light = Set(['top', 'middle', 'bottom'])
    color = Regex(r'(red|yellow|blue|orange|purple|...)')
    times = Set(['once', 'twice', 'three times']) | Regex(r'\d+ times')
    one_parse = action + light + Optional(times) + color
    GOAL = OneOrMore(one_parse)

    @staticmethod
    def test():
        parser = RobustParser((LightGrammar()))
        sents = [
            "set my top light to red",
            "set my top light to red and change middle light to yellow",
            "set my top light to red and change middle light to yellow and "
            "flash bottom light twice in blue"
        ]
        for sent in sents:
            tree, result = parser.parse(sent)
            assert result.one_parse[0].color == 'red'

            print '"%s"' % sent
            print "parse tree:"
            print tree
            print "parse result:"
            print result
            print

Dependencies

None. Parsetron is a single parsetron.py file.

Parsetron is inspired by pyparsing.

Grammar Modules

Parsetron supports modularized grammars: each grammar focuses on an individual small domain and can be imported via, for instance:

from parsetron.grammars.colors import ColorsGrammar

class YourCustomizedGrammar(Grammar):
    color = ColorsGrammar.GOAL

You are welcome to contribute your own grammar here (under parsetron.grammars). Send us a pull request!

Development

  1. fork this repository

  2. install dev-specific packages:

    pip install -r requirements.txt
    
  3. then make your changes and follow the Makefile.

TODO

  • [ ] Python 3 compatible
  • [ ] Unicode support