en-pos

A better English POS tagger written in JavaScript

Installation and usage

Install via NPM:

npm i --save en-pos

How to use

const Tag = require("en-pos").Tag;
var tags = new Tag(["this","is","my","sentence"])
.initial() // initial dictionary and pattern based tagging
.smooth() // further context based smoothing
.tags;
console.log(tags);
// ["DT","VBZ","PRP$","NN"]

Annotation Specification

Annotation	Name	Example
NN	Noun	dog man
NNS	Plural noun	dogs men
NNP	Proper noun	London Alex
NNPS	Plural proper noun	Smiths
VB	Base form verb	be
VBP	Present form verb	throw
VBZ	Present form (3rd person)	throws
VBG	Gerund form verb	throwing
VBD	Past tense verb	threw
VBN	Past participle verb	thrown
MD	Modal verb	can shall will may must ought
JJ	Adjective	big fast
JJR	Comparative adjective	bigger
JJS	Superlative adjective	biggest
RB	Adverb	not quickly closely
RBR	Comparative adverb	less-closely faster
RBS	Superlative adverb	fastest
DT	Determiner	the a some both
PDT	Predeterminer	all quite
PRP	Personal Pronoun	I you he she
PRP$	Possessive Pronoun	I you he she
POS	Possessive ending	's
IN	Preposition	of by in
PR	Particle	up off
TO	to	to
WDT	Wh-determiner	which that whatever whichever
WP	Wh-pronoun	who whoever whom what
WP$	Wh-possessive	whose
WRB	Wh-adverb	how where
EX	Expletive there	there
CC	Coordinating conjugation	& and nor or
CD	Cardinal Numbers	1 7 77 one
LS	List item marker	1 B C One
UH	Interjection	ah oh oops
FW	Foreign Words	viva mon toujours
,	Comma	,
:	Mid-sent punct	: ; ...
.	Sent-final punct.	. ! ?
(	Left parenthesis	) } ]
)	Right parenthesis	( { [
#	Pound sign	#
$	Currency symbols	$ € £ ¥
SYM	Other symbols	+ * / < >
EM	Emojis & emoticons	:) ❤

Accuracy and performance

TL:DR;

• When smoothing is enabled: 96.43% accuracy (processing 132K tokens in 38 seconds)
• When smoothing is disabled: 94.4% accuracy (processing 132K tokens in 3 seconds)

---

As of 25 Jan 2017, this library scored 96.43% at the Penn Treebank test (0.3% away from being a state of the art tagger).

Being written in JavaScript, I think it's safe to say that this is the most accurate JavaScript POS tagger, since the only JS library I know of is pos-js which when I tested on the same treebank scored 87.8%, though it was faster than my implementation when smoothing is enabled.

However, if performance is what's you're after rather than accuracy, then you have the option to disable smoothing in this library and this will marginally increase performance making this library even faster than pos-js but with far better accuracy (94.4%).

Building from source and testing

• Build: tsc (requires typescript)
• Test: node test/test.ts

Credits

• This project is an optimization and (almost complete) re-writing of Compendium's POS tagger.
• Compendium's tagger itself was based on fasttag.
• Fasttag is based on Eric Brill's POS tagger.