textstem

Project Status: Active - The project has reached a stable, usable state and is being actively developed. Build Status Coverage Status

textstem is a tool-set for stemming and lemmatizing words. Stemming is a process that removes affixes. Lemmatization is the process of grouping inflected forms together as a single base form.

Table of Contents

Functions

The main functions, task category, & descriptions are summarized in the table below:

Function Task Description
stem_words stemming Stem words
stem_strings stemming Stem strings
lemmatize_words lemmatizing Lemmatize words
lemmatize_strings lemmatizing Lemmatize strings
make_lemma_dictionary_words lemmatizing Generate a dictionary of lemmas for a text

Installation

To download the development version of textstem:

Download the zip ball or tar ball, decompress and run R CMD INSTALL on it, or use the pacman package to install the development version:

if (!require("pacman")) install.packages("pacman")
pacman::p_load_gh("trinker/textstem")

Contact

You are welcome to:
- submit suggestions and bug-reports at: https://github.com/trinker/textstem/issues
- send a pull request on: https://github.com/trinker/textstem/
- compose a friendly e-mail to: tyler.rinker@gmail.com

Examples

The following examples demonstrate some of the functionality of textstem.

Load the Tools/Data

if (!require("pacman")) install.packages("pacman")
pacman::p_load(textstem, dplyr)

data(presidential_debates_2012)

Stemming Versus Lemmatizing

Before moving into the meat these two examples let’s highlight the difference between stemming and lemmatizing.

“Drive” Stemming vs. Lemmatizing

dw <- c('driver', 'drive', 'drove', 'driven', 'drives', 'driving')

stem_words(dw)

## [1] "driver" "drive"  "drove"  "driven" "drive"  "drive"

lemmatize_words(dw)

## [1] "driver" "drive"  "drive"  "drive"  "drive"  "drive"

“Be” Stemming vs. Lemmatizing

bw <- c('are', 'am', 'being', 'been', 'be')

stem_words(bw)

## [1] "ar"   "am"   "be"   "been" "be"

lemmatize_words(bw)

## [1] "be" "be" "be" "be" "be"

Stemming

Stemming is the act of removing inflections from a word not necessarily “identical to the morphological root of the word” (wikipedia). Below I show stemming of several small strings.

y <- c(
    'the dirtier dog has eaten the pies',
    'that shameful pooch is tricky and sneaky',
    "He opened and then reopened the food bag",
    'There are skies of blue and red roses too!',
    NA,
    "The doggies, well they aren't joyfully running.",
     "The daddies are coming over...",
    "This is 34.546 above"
)
stem_strings(y)

## [1] "the dirtier dog ha eaten the pi"          
## [2] "that shame pooch i tricki and sneaki"     
## [3] "He open and then reopen the food bag"     
## [4] "There ar ski of blue and red rose too!"   
## [5] NA                                         
## [6] "The doggi, well thei aren't joyfulli run."
## [7] "The daddi ar come over..."                
## [8] "Thi i 34.546 abov"

Lemmatizing

Default Lemma Dictionary

Lemmatizing is the “grouping together the inflected forms of a word so they can be analysed as a single item” (wikipedia). In the example below I reduce the strings to their lemma form. lemmatize_strings uses a lookup dictionary. The default uses Mechura’s (2016) English lemmatization list available from the lexicon package. The make_lemma_dictionary function contains two additional engines for generating a lemma lookup table for use in lemmatize_strings.

y <- c(
    'the dirtier dog has eaten the pies',
    'that shameful pooch is tricky and sneaky',
    "He opened and then reopened the food bag",
    'There are skies of blue and red roses too!',
    NA,
    "The doggies, well they aren't joyfully running.",
     "The daddies are coming over...",
    "This is 34.546 above"
)
lemmatize_strings(y)

## [1] "the dirty dog have eat the pie"           
## [2] "that shameful pooch be tricky and sneaky" 
## [3] "He open and then reopen the food bag"     
## [4] "There be sky of blue and red rose too!"   
## [5] NA                                         
## [6] "The doggy, good they aren't joyfully run."
## [7] "The daddy be come over..."                
## [8] "This be 34.546 above"

Hunspell Lemma Dictionary

This lemmatization uses the hunspell package to generate lemmas.

lemma_dictionary_hs <- make_lemma_dictionary(y, engine = 'hunspell')
lemmatize_strings(y, dictionary = lemma_dictionary_hs)

## [1] "the dirty dog ha eat the pie"               
## [2] "that shameful pooch i tricky and sneaky"    
## [3] "He open and then reopen the food bag"       
## [4] "There are sky of blue and re rose too!"     
## [5] NA                                           
## [6] "The doggy, well they aren't joyful running."
## [7] "The daddy are come over..."                 
## [8] "This i 34.546 above"

koRpus Lemma Dictionary

This lemmatization uses the koRpus package and the TreeTagger program to generate lemmas. You’ll have to get TreeTagger set up, preferably in your machine’s root directory.

lemma_dictionary_tt <- make_lemma_dictionary(y, engine = 'treetagger')
lemmatize_strings(y, lemma_dictionary_tt)

## [1] "the dirty dog have eat the pie"           
## [2] "that shameful pooch be tricky and sneaky" 
## [3] "He open and then reopen the food bag"     
## [4] "There be sky of blue and red rose too!"   
## [5] NA                                         
## [6] "The doggy, well they aren't joyfully run."
## [7] "The daddy be come over..."                
## [8] "This be 34.546 above"

Lemmatization Speed

It’s pretty fast too. Observe:

tic <- Sys.time()

presidential_debates_2012$dialogue %>%
    lemmatize_strings() %>%
    head()

## [1] "We'll talk about specifically about health care in a moment."                            
## [2] "But what do you support the voucher system, Governor?"                                   
## [3] "What I support be no change for current retiree and near retiree to Medicare."           
## [4] "And the president support take dollar seven hundred sixteen billion out of that program."
## [5] "And what about the voucher?"                                                             
## [6] "So that's that's numb one."

(toc <- Sys.time() - tic)

## Time difference of 0.8516021 secs

That’s 2,912 rows of text, or 42,708 words, in 0.85 seconds.

Combine With Other Text Tools

This example shows how stemming/lemmatizing might be complemented by other text tools such as replace_contraction from the textclean package.

library(textclean)

'aren\'t' %>% 
    lemmatize_strings()

## [1] "aren't"

'aren\'t' %>% 
    textclean::replace_contraction() %>%
    lemmatize_strings()

## [1] "be not"