BUG FIXES
replace_emoticon replaced emoticon-like substrings within actual words.
Spotted thanks to Carolyn Challoner; see issue #46.
replace_number failed if the number pattern contained two leading decimals
or hyphens. Spotted thanks to Stefano De Sabbata; see issue #60.
replace_word_elongation failed for repeating of the same character but of
different case (e.g., replace_word_elongation("Ooo") resulted in NA. This
has been corrected. Additionally, the elongation.search.pattern defined as
"(?i)(?:^|\\b)\\w*([a-z])(?:\\1{2,})\\w*($|\\b)" has been moved exterally, to
a parameter, allowing the user to alter this pattern if desired. Spotted
thanks to Stefano De Sabbata; see issue #59.
NEW FEATURES
replace_misspelling added as a way to replace misspelled words with their
most likely replacement using hunspell in the backend. Suggested by Surin
Space; see issue #39.
as_ordinal added as a convenience wrapper for english::ordinal that
takes integers and converts them to ordinal form.
%like% added as an binary operator similar to SQL's LIKE.
MINOR FEATURES
fix_mdyyyy added to correct dates in the form of m/d/yyyy to yyyy-mm-dd.IMPROVEMENTS
replace_html pics up the ability to replace "«" & "»" with ASCII
equivalents "<<" & ">>". Suggested by Ilya Shutov; see issue #48.
All internal calls to grepl() now have perl = TRUE added as this is
generally a speed up. Suggested by Kyle Haynes (see #51).
CHANGES
filter_element() and filter_row() have been deprecated for a few years.Version update to comply with changes in the glue package's API.
BUG FIXES
fgsub had a bug in which the the original pattern in fgsub matches the
location in the string but when the replacement occurs this was done on the
entire string rather than the location of the first pattern match. This
means the extracted string was used as a search and might be found in places
other than the original location (e.g., a leading boundary in '^T' replaced
with '__' may have led to '__he __itle' rather than '__he Title' as expected
in the string 'The Title'). See #35 for details. The fix will add some time
to the computation but is safer.NEW FEATURES
replace_to/replace_from added to remove from/to begin/end of string to/from
a character(s).
The following replacement functions were added to provide remediation for
problems found in check_text: replace_email, replace_hash,
replace_tag, & replace_url.
MINOR FEATURES
check_text picks up a checks and n argument. The former allows the user
to specify which checks to conduct. The latter allows the user to truncate the
output to n number of elements with a closing ...[truncated].... This makes
the function more useful and the code easier to maintain.IMPROVEMENTS
replace_non_ascii did not replace all non-ASCII characters. This has been
fixed by an explicit replacement of '[^ -~]+' which are all non-ASCII characters.
See issue #34 for details.Maintenance release to bring package up to date with the lexicon package API changes.
NEW FEATURES
match_tokens added to find all the tokens that match a regex(es) within a
given text vector. This useful when combined with the replace_tokens
function.
Fixed versions of drop_element/keep_element added to allow for dropping
elements specified by a known vector rather than a regex.
The collapse and glue functions from the glue package are reexported
for easy string manipulation.
replace_date added for normalizing dates.
replace_time added for normalizing time stamps.
replace_money added for normalizing money references.
mgsub picks up a safe argument using the mgsub package as the backend.
In addition mgsub_regex_safe added to make the usage explicit. The safe mode
comes at the cost of speed.
IMPROVEMENTS
replace_names drops the replacement of
c('An', 'To', 'Oh', 'So', 'Do', 'He', 'Ha', 'In', 'Pa', 'Un') which are
likely words and not names.
replace_html picks ups some additional symbol replacements including:
c("™", "“", "”", "‘", "’", "•", "·", "⋅", "–", "—", "≠", "½", "¼", "¾", "°", "←", "→", "…").
NEW FEATURES
replace_kern added to replace a form of informal emphasis in which the
writer takes words >2 letters long, capitalizes the entire word, and places
spaces in between each letter. This was contributed by Stack Overflow's
@ctwheels: https://stackoverflow.com/a/47438305/1000343.
replace_internet_slang added to replace Internet acronyms and abbreviations
with machine friendly word equivalents.
replace_word_elongation added to replace word elongations (a.k.a. "word
lengthening") with the most likely normalized word form. See
http://www.aclweb.org/anthology/D11-105 for details.
fgsub added for the ability to match, extract, operate a function over the
extracted strings, & replace the original matches with the extracted strings.
This performs similar functionality to gsubfn::gsubfn but is less powerful.
For more powerful needs see the gsubfn package.
BUG FIXES
replace_grade did not use fixed = TRUE for its call to mgsub. This could
result in the plus signs being interpreted as meta-characters. This has been
corrected.NEW FEATURES
replace_names added to remove/replace common first and last names from text
data.
make_plural added to make a vector of singular noun forms plural.
replace_emoji and replace_emoji_identifier added for replacing emojis with
text or an identifier token for use in the sentimentr package.
MINOR FEATURES
mgsub_regex and mgsub_fixed to provide wrappers for mgsub that makes
their use apparent without setting the fixed command.
replace_curly_quote added to replace curly quotes with straight versions.
IMPROVEMENTS
replace_non_ascii now uses stringi::stri_trans_general to coerce more
non-ASCII characters to ASCII format.
check_text now checks for HTML characters/tags. Thanks to @Peter Gensler
for suggesting this (see issue #15).
CHANGES
filter_ functions deprecated in favor of drop_/keep_ versions of filter
functions. This was change was to address the opposite meaning that dplyr's
filter has, which retains rows matching a pattern be default.BUG FIXES
replace_tokens added to complement mgsub for times when the user wants to
replace fixed tokens with a single value or remove them entirely. This yields
an optimized solution that is much faster than mgsub.CHANGES
mgusb no longer uses trim = TRUE by default.BUG FIXES
check_text reported to use replace_incomplete rather than
add_missing_endmark when endmark is missing.NEW FEATURES
The replace_emoticon, replace_grade and replace_rating functions have
been moved from the sentimentr package to textclean as these are
cleaning functions. This makes the functions more modular and generalizable
to all types of text cleaning. These functions are still imported and
exported by sentimentr.
replace_html added to remove html tags and repalce symbols with appropriate
ASCII symbols.
add_missing_endmarks added to detect missing endmarks and replace with the
desired symbol.
IMPROVEMENTS
replace_number now uses the english package making it faster and more
maintainable. In addition, the function now handles decimal places as well.BUG FIXES
check_text reported NA as non-ASCII. This has been fixed.NEW FEATURES
check_text added to report on potential problems in a text vector.
replace_ordinal added to replace ordinal numbers (e.g., 1st) with word
representation (e.g., first).
swap added to swap two patterns simultaneously.
filter_element added to exclude matching elements from a vector.
This package is a collection of tools to clean and process text. Many of these tools have been taken from the qdap package and revamped to be more intuitive, better named, and faster.