Skip to content
Lemmatizer for text in English. Inspired by Python's nltk.corpus.reader.wordnet.morphy
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.


Type Name Latest commit message Commit time
Failed to load latest commit information.
lib Revert "chenged = to ==" Apr 27, 2019
spec abbreviation resolution added Apr 18, 2019
.gitignore initial commit Oct 28, 2012
Gemfile initial commit Oct 28, 2012
LICENSE.txt code organization refined by Vladimir Ivic Nov 3, 2013 README modified Apr 18, 2019
Rakefile now user can supply lex data at runtime Feb 15, 2019
lemmatizer.gemspec Update specs, remove encoding headers Sep 13, 2015


Lemmatizer for text in English. Inspired by Python's nltk.corpus.reader.wordnet.morphy package.

Based on code posted by mtbr at his blog entry WordNet-based lemmatizer

Version 0.2 has added functionality to add user supplied data at runtime


sudo gem install lemmatizer


require "lemmatizer"
lem =
p lem.lemma("dogs",    :noun ) # => "dog"
p lem.lemma("hired",   :verb ) # => "hire"
p lem.lemma("hotter",  :adj  ) # => "hot"
p lem.lemma("better",  :adv  ) # => "well"
# when part-of-speech symbol is not specified as the second argument, 
# lemmatizer tries :verb, :noun, :adj, and :adv one by one in this order.
p lem.lemma("fired")           # => "fire"
p lem.lemma("slow")            # => "slow"


# Lemmatizer leaves alone words that its dictionary does not contain.
# This keeps proper names such as "James" intact.
p lem.lemma("MacBooks", :noun) # => "MacBooks" 
# If an inflected form is included as a lemma in the word index,
# lemmatizer may not give an expected result.
p lem.lemma("higher", :adj) # => "higher" not "high"!

# The above has to happen because "higher" is itself an entry word listed in dict/index.adj .
# To fix this, modify the original dict directly (lib/dict/index.{noun|verb|adj|adv}) 
# or supply with custom dict files (recommended).

Supplying with user dict

# You can supply custom dict files consisting of lines in the format of <pos>\s+<form>\s+<lemma>.
# The data in user supplied files overrides the preset data. Here's the sample. 

# --- sample.dict1.txt (don't include hash symbol on the left) ---
# adj   higher   high
# adj   highest  high
# noun  MacBooks MacBook
# ---------------------------------------------------------------

lem ="sample.dict1.txt")

p lem.lemma("higher", :adj)     # => "high"
p lem.lemma("highest", :adj)    # => "high"
p lem.lemma("MacBooks", :noun)  # => "MacBook"

# The argument to can be either of the following:
# 1) a path string to a dict file (e.g. "/path/to/dict.txt")
# 2) an array of paths to dict files (e.g. ["./dict/noun.txt", "./dict/verb.txt"])

Resolving abbreviations

# You can use 'abbr' tag in user dicts to resolve abbreviations in text.

# --- sample.dict2.txt (don't include hash symbol on the left) ---
# abbr  utexas   University of Texas
# abbr  mit      Massachusetts Institute of Technology
# ---------------------------------------------------------------

# <NOTE>
# 1. Expressions on the right (substitutes) can contain white spaces, 
#    while expressions in the middle (words to be replaced) cannot.
# 2. Double/Single quotations could be used with substitute expressions,
#    but not with original expressions.

lem ="sample.dict2.txt")

p lem.lemma("utexas", :abbr) # => "University of Texas"
p lem.lemma("mit", :abbr)    # => "Massachusetts Institute of Technology"


Thanks for assistance and contributions:


Licensed under the MIT license.

You can’t perform that action at this time.