parse-english

Copyright (C) 2011-2017 mailto:[email protected]

About

parse-english is a minimum viable English parser implemented in LexYacc. It parses in parallel all possible interpretations of an English sentence accepted by a grammar and generates abstract syntax trees for successful parses. The algorithm is completely deterministic. No training data is required.

See old version here: NatLang

A Motivating Example

input:

the quick brown fox jumps over the lazy dog.

output:

Usage

cd ./demo/0_parse-english_full_nlp
./demo.sh "the quick brown fox jumps over the lazy dog"

Switch	Description
-e SENTENCE	input sentence
-l	Lisp mode
-g	graph mode (slow for deep trees)
-d	dot mode
-x	extract ontology mode
-q	quiet mode
-m	memory debug
-n	indent lisp

Requirements

Unix tools and 3rd party components (accessible from $PATH):

gcc flex bison

Supported Features

• Parallel reentrant parsing
• Lisp / graph / dot output (multiple trees)

Supported Grammar Syntaxes

• Present tense
• Progressive tense
• Future tense
• Past tense
• Past perfect tense
• Passive voice
• Questions
• Conditionals
• Imperitive mood
• Comparisons

Limitations

• Hard coded grammar & vocabulary.
• A brute force algorithm tries all supported interpretations of a sentence. This is slow for long sentences.
• BNF rules are suitable for specifying constituent-based phrase structure grammars, but are a poor fit for expressing non-local dependencies.

Make Targets

target	action
all	make binaries
test	all + run tests
pure	test + use valgrind to check for memory leaks
dot	test + generate .png graph for tests
lint	use cppcheck to perform static analysis on .cpp files
doc	use doxygen to generate documentation
xml	test + generate .xml for tests
import	test + use ticpp to serialize-to/deserialize-from xml
clean	remove all intermediate files

References

"Part-of-speech tagging"

http://en.wikipedia.org/wiki/Part-of-speech_tagging

"Princeton WordNet"

http://wordnet.princeton.edu/

"Syntactic Theory: A Unified Approach"

ISBN: 0340706104

"Enju - A fast, accurate, and deep parser for English"

http://www.nactem.ac.uk/enju/

Keywords

Natural Language Processing, English parser, Yacc, BNF