en-parse

English Parse

English language syntactic dependency parser.

important This project is in a very early stage, it's rather an experiment than a real a project. Using it in production may hurt your feelings.

Installation and Usage

npm i --save en-parse

const parser = require("en-parse").default;
var tokens = ["The", "problem", "is", "that", "this", "has", "never", "been", "tried", "."];
var tags = ["DT","NN","VBZ","IN","DT","VBZ","RB","VBN","VBD","PUNCT"];
let parsed = parser(["PRP","VBP","TO","VBP"],["i","like","to","parse"]);

NOTE: tokens should be normalized (contractions, abbreviations and such all resolved to it's normal form).

The parsed variable is now an array like this one:

[
  {
    "label": "MWE",
    "type": "NP",
    "master": 1
  },
  {
    "label": "NSUBJ",
    "type": "NP",
    "master": 2
  },
  {
    "label": "ROOT",
    "type": "VP",
    "master": -1
  },
  {
    "label": "COMPMARK",
    "type": "PP",
    "master": 5
  },
  {
    "label": "NSUBJ",
    "type": "NP",
    "master": 5
  },
  {
    "label": "XCOMP",
    "type": "VP",
    "master": 2
  },
  {
    "label": "ADVMOD",
    "type": "ADV",
    "master": 8
  },
  {
    "label": "AUX",
    "type": "VBN",
    "master": 8
  },
  {
    "label": "ADVCL",
    "type": "VP",
    "master": 5
  },
  {
    "label": "PUNCT",
    "type": "PUNCT",
    "master": 2
  }
]

Annotation Specs:

Core Nominal Arguments

Annotation	Name	Example
NSUBJ	Nominal Subject	I like you
DOBJ	Direct Object	I like you
IOBJ	Indirect Object	she gave me a book
NSUBJPASS	Nominal Subject (passive)	I have been given a chance
ATTR	Attribute	This is awesome
CCOMP	Clausal Complement	The boss ordered to dig the hole
XCOMP	Open Clausal Complement	The boss told us to dig
OBL	Oblique Object	Give the toys to the children
EXPL	Expletive	I have been there
ADVCL	Adverb Clause Modifier	We were walking as the rain was falling
ADVMOD	Adverbial modifier	I don't eat genetically modified food.
DISCOURSE	Discourse	I like that :)
VPRT	Verbal Particle	I switched the TV off
AUX	Auxiliary	I am going to buy a new computer
AUXPASS	Auxiliary (passive)	I have been marked
COMPMARK	Complement Marker	I will do it if I like it
ADVMARK	Adverb Marker	I did it after I was convinced
NOMD	Nominal Modifier	MSNBC news had little effect on the market
NUMMOD	Numeric Modifier	I ate 2 eggs this morning
ACL	Clausal Modifier of a Noun	I saw the man you love
AMOD	Adjectival Modifier	you were good to him
DET	Determiner	which book you prefer?
CASE	Case Marking	I went to Rachel
PUNCT	Punctuation	Guys, keep calm!
CC	Coordinating Conjunction	Mathew and Alex
INTERJ	Interjection	Pass me the sugar, please
DEP	Unrecognized dependency	how what

Why?

• Current solutions uses statistical machine learning models which makes it 5 times slower (if not worse) than a rule based implementation like this one.
• This is a javascript implementation, so it can be used for an in-browser natural language processing.
• Also because it's written in javascript it might be preferred in smart bots to reduce the complexity of having your bot written in node.js and the natural language processor written in Python or C++.

Credits

The idea of this parser was originally featured as a small experiment in Xav Ulflander's Compendium.