wordpos is a set of fast part-of-speech (POS) utilities for Node.js using fast lookup in the WordNet database.

Version 1.x is a major update with no direct dependence on natural's WordNet module, with support for Promises, and roughly 5x speed improvement over previous version.

CAUTION The WordNet database wordnet-db comprises 155,287 words (3.0 numbers) which uncompress to over 30 MB of data in several unbrowserify-able files. It is not meant for the browser environment.

Node.js:

var WordPOS = require('wordpos'),
    wordpos = new WordPOS();

wordpos.getAdjectives('The angry bear chased the frightened little squirrel.', function(result){
    console.log(result);
});
// [ 'little', 'angry', 'frightened' ]

wordpos.isAdjective('awesome', function(result){
    console.log(result);
});
// true 'awesome'

Command-line: (see CLI for full command list)

$ wordpos def git
git
  n: a person who is deemed to be despicable or contemptible; "only a rotter would do that"; "kill the rat"; "throw the bum out"; "you cowardly little pukes!"; "the British call a contemptible person a `git'"  

$ wordpos def git | wordpos get --adj
# Adjective 6:
despicable
contemptible
bum
cowardly
little
British

 npm install -g wordpos

To run test: (or just: npm test)

npm install -g mocha
mocha test

WordPOS.defaults = {
  /**
   * enable profiling, time in msec returned as last argument in callback
   */
  profile: false,

  /**
   * if true, exclude standard stopwords.
   * if array, stopwords to exclude, eg, ['all','of','this',...]
   * if false, do not filter any stopwords.
   */
  stopwords: true
};

To override, pass an options hash to the constructor. With the profile option, most callbacks receive a last argument that is the execution time in msec of the call.

    wordpos = new WordPOS({profile: true});
    wordpos.isAdjective('fast', console.log);
    // true 'fast' 29

Please note: all API are async since the underlying WordNet library is async.

Get part-of-speech from text. callback(results) receives an array of words for specified POS, or a hash for getPOS():

wordpos.getPOS(text, callback) -- callback receives a result object:
    {
      nouns:[],       Array of words that are nouns
      verbs:[],       Array of words that are verbs
      adjectives:[],  Array of words that are adjectives
      adverbs:[],     Array of words that are adverbs
      rest:[]         Array of words that are not in dict or could not be categorized as a POS
    }
    Note: a word may appear in multiple POS (eg, 'great' is both a noun and an adjective)

If you're only interested in a certain POS (say, adjectives), using the particular getX() is faster than getPOS() which looks up the word in all index files. stopwords are stripped out from text before lookup.

If text is an array, all words are looked-up -- no deduplication, stopword filtering or tokenization is applied.

getX() functions return a Promise.

Example:

wordpos.getNouns('The angry bear chased the frightened little squirrel.', console.log)
// [ 'bear', 'squirrel', 'little', 'chased' ]

wordpos.getPOS('The angry bear chased the frightened little squirrel.', console.log)
// output:
  {
    nouns: [ 'bear', 'squirrel', 'little', 'chased' ],
    verbs: [ 'bear' ],
    adjectives: [ 'little', 'angry', 'frightened' ],
    adverbs: [ 'little' ],
    rest: [ 'the' ]
  }

This has no relation to correct grammar of given sentence, where here only 'bear' and 'squirrel' would be considered nouns.

Determine if word is a particular POS. callback(result, word) receives true/false as first argument and the looked-up word as the second argument. The resolved Promise receives true/false.

Examples:

wordpos.isVerb('fish', console.log);
// true 'fish'

wordpos.isNoun('fish', console.log);
// true 'fish'

wordpos.isAdjective('fishy', console.log);
// true 'fishy'

wordpos.isAdverb('fishly', console.log);
// false 'fishly'

Get complete definition object for word. The lookupX() variants can be faster if you already know the POS of the word. Signature of the callback is callback(result, word) where result is an array of lookup object(s).

Example:

wordpos.lookupAdjective('awesome', console.log);
// output:
[ { synsetOffset: 1285602,
    lexFilenum: 0,
    lexName: 'adj.all',
    pos: 's',
    wCnt: 5,
    lemma: 'amazing',
    synonyms: [ 'amazing', 'awe-inspiring', 'awesome', 'awful', 'awing' ],
    lexId: '0',
    ptrs: [],
    gloss: 'inspiring awe or admiration or wonder; [...] awing majesty, so vast, so high, so silent"  '
    def: 'inspiring awe or admiration or wonder',     
    ...
} ], 'awesome'

In this case only one lookup was found, but there could be several.

Version 1.1 adds the lexName parameter, which maps the lexFilenum to one of 45 lexicographer domains.

Version 1.1 introduces the seek method to lookup a record directly from the synsetOffset for a given POS. Unlike other methods, callback (if provided) receives (err, result) arguments.

Examples:

wordpos.seek(1285602, 'a').then(console.log)
// same result as wordpos.lookupAdjective('awesome', console.log);

Get random word(s). (Introduced in version 0.1.10) callback(results, startsWith) receives array of random words and the startsWith option, if one was given. options, if given, is:

{
  startsWith : <string> -- get random words starting with this
  count : <number> -- number of words to return (default = 1)
}

Examples:

wordpos.rand(console.log)
// ['wulfila'] ''

wordpos.randNoun(console.log)
// ['bamboo_palm'] ''

wordpos.rand({starstWith: 'foo'}, console.log)
// ['foot'] 'foo'

wordpos.randVerb({starstWith: 'bar', count: 3}, console.log)
// ['barge', 'barf', 'barter_away'] 'bar'

wordpos.rand({starsWith: 'zzz'}, console.log)
// [] 'zzz'

Note on performance: random lookups could involve heavy disk reads. It is better to use the count option to get words in batches. This may benefit from the cached reads of similarly keyed entries as well as shared open/close of the index files.

Getting random POS (randNoun(), etc.) is generally faster than rand(), which may look at multiple POS files until count requirement is met.

Returns tokenized array of words in text, less duplicates and stopwords. This method is called on all getX() calls internally.

Access to the wordnet-db object containing the dictionary & index files.

Access the array of stopwords.

As of v1.0, all get, is, rand, and lookup methods return a standard ES6 Promise.

wordpos.isVerb('fish').then(console.log);
// true

Compound, with error handler:

wordpos.isVerb('fish')
  .then(console.log)
  .then(doSomethingElse)
  .catch(console.error);

Callbacks, if given, are executed before the Promise is resolved.

wordpos.isVerb('fish', console.log)
  .then(console.log)
  .catch(console.error);
// true 'fish' 13
// true

Note that callback receives full arguments (including profile, if enabled), while the Promise receives only the result of the call. Also, beware that exceptions in the callback will result in the Promise being rejected and caught by catch(), if provided.

Version 0.1.4 introduces fastIndex option. This uses a secondary index on the index files and is much faster. It is on by default. Secondary index files are generated at install time and placed in the same directory as WNdb.path. Details can be found in tools/stat.js.

Fast index improves performance 30x over Natural's native methods. See blog article Optimizing WordPos.

As of version 1.0, fast index is always on and cannot be turned off.

For CLI usage and examples, see bin/README.

See bench/README.

1.2.0

• Fix new Buffer() deprecation warning.
• Fix npm audit vulnerabilities

1.1.6

• Fix #25 rand().then with no args

1.1.5

• rollback 1.1.4 changes. Fix is made in wordnet-db.

1.1.4

• temporary fix for #19 issue with npm@5

1.1.2

• Fix DeprecationWarning for node 7.x (1.1.1)
• Fix occasional error for large offsets during seek

1.1.0

• added seek() method
• added lexName property

1.0.1

• Removed npm dependency on Natural. Certain modules are included in /lib.
• Add support for ES6 Promises.
• Improved data file reads for up to 5x performance increase compared to previous version.
• Tests are now mocha-based with chai assert interface.

0.1.16

• Changed dependency to wordnet-db (renamed from WNdb)

0.1.15

• Added syn (synonym) and exp (example) CLI commands.
• Fixed rand CLI command when no start word given.
• Removed -N, --num CLI option. Use wordpos rand [N] to get N random numbers.
• Changed CLI option -s to -w (include stopwords).

0.1.13

• Fix crlf issue for command-line script

0.1.12

• fix stopwords not getting excluded when running with CLI
• added 'stopwords' CLI command to show list of stopwords
• CLI option --stopword now renamed to --withStopwords

0.1.10

• rand functionality added

0.1.6

• added command line tool

0.1.4

• added fast index

(The MIT License)

Copyright (c) 2012, 2014, 2016 [email protected]