Stemming with SnowballC vs hunspell

stemming-snowballc-hunspell.R

#ref: https://github.com/juliasilge/tidytext/issues/17

library(dplyr)
library(hunspell)
library(SnowballC)

w <- tibble(
  palabras = c(
    "celebra",
    "celebré",
    "celebraré",
    "celebramos",
    "celebrado",
    "celebraremos",
    "celebró"
  )
)

# stemming Spanish with SnowballC
df1 <- w %>% 
  mutate(stem = wordStem(palabras, language = "spanish"))
df1

# A tibble: 7 x 2
#   palabras     stem  
#   <chr>        <chr> 
# 1 celebra      celebr
# 2 celebré      celebr
# 3 celebraré    celebr
# 4 celebramos   celebr
# 5 celebrado    celebr
# 6 celebraremos celebr
# 7 celebró      celebr

# and now with hunspell
df2 <- w %>% 
  mutate(
    stem = hunspell_stem(palabras, dict = dictionary("es_PE"))
  ) %>% 
  unnest(stem)
df2

# A tibble: 8 x 2
#   palabras     stem     
#   <chr>        <chr>    
# 1 celebra      celebrar 
# 2 celebré      celebrar 
# 3 celebraré    celebrar 
# 4 celebramos   celebrar 
# 5 celebrado    celebrado
# 6 celebrado    celebrar 
# 7 celebraremos celebrar 
# 8 celebró      celebrar

# Somehow hunspell makes a bit more sense