R: Hunspell Spell Checking and Morphological Analysis
hunspell
R Documentation
Hunspell Spell Checking and Morphological Analysis
Description
The hunspell function is a high-level wrapper for finding spelling
errors within a text document. It takes a character vector with text (plain,
latex or man format), parses out the words and returns a list with
incorrect words for each line. It effectively combines of hunspell_parse
with hunspell_check in a single step. Other functions in the package
operate on individual words, see details.
input format; supported parsers are text, latex or man
dict
dictionary language, see details
ignore
character vector with additional approved words added to the dictionary
words
character vector with individual words to spell check
Details
Hunspell uses a special dictionary format that defines which stems and affixes are
valid in a given language. The hunspell_analyze function shows how a
word breaks down into a valid stem plus affix. The hunspell_stem
function is similar but only returns valid stems for a given word. Stemming can be
used to summarize text (e.g in a wordcloud). The hunspell_check function
takes a vector of individual words and tests each one for correctness. Finally
hunspell_suggest is used to suggest correct alternatives for each
(incorrect) input word.
Because spell checking is usually done on a document, the package includes some
parsers to extract words from various common formats. With hunspell_parse
we can parse plain-text, latex and man format. R also has a few built-in parsers
such as RdTextFilter and
SweaveTeXFilter, see also
?aspell.
The package searches for dictionaries in the working directory as well as in the
standard system locations. Additional search paths can be specified by setting
the DICPATH environment variable. A US English dictionary (en_US) is
included with the package; other dictionaries need to be installed by the system.
Most operating systems already include compatible dictionaries with names such as
hunspell-en-gb or
myspell-en-gb.
To manually install dictionaries, download the .aff and .dic file
from an OpenOffice mirror
or bundle
and copy them to ~/Library/Spelling or a custom directory specified in DICPATH.
Alternatively you can pass the entire path to the .dic file as the dict
parameter.
Note that hunspell uses iconv to convert input text to
the encoding used by the dictionary. This will fail if text contains characters
which are unsupported by that particular encoding. For this reason UTF-8 dictionaries
are preferable over legacy 8bit dictionaries Several UTF8 dictionaries are
available from Github.
Examples
# Check individual words
words <- c("beer", "wiskey", "wine")
correct <- hunspell_check(words)
print(correct)
# Find suggestions for incorrect words
hunspell_suggest(words[!correct])
# Extract incorrect from a piece of text
bad <- hunspell("spell checkers are not neccessairy for langauge ninja's")
print(bad[[1]])
hunspell_suggest(bad[[1]])
# Stemming
words <- c("love", "loving", "lovingly", "loved", "lover", "lovely", "love")
hunspell_stem(words)
hunspell_analyze(words)
# Check an entire latex document
setwd(tempdir())
download.file("http://arxiv.org/e-print/1406.4806v1", "1406.4806v1.tar.gz", mode = "wb")
untar("1406.4806v1.tar.gz")
text <- readLines("content.tex", warn = FALSE)
bad_words <- hunspell(text, format = "latex")
sort(unique(unlist(bad_words)))
# Summarize text by stems (e.g. for wordcloud)
allwords <- hunspell_parse(text, format = "latex")
stems <- unlist(hunspell_stem(unlist(allwords)))
words <- head(sort(table(stems), decreasing = TRUE), 200)