keyToEnglish

The keyToEnglish package provides a function to create an easy-to-remember hash of a field by using a traditional hashing function, and then mapping substrings of the hash to a list of words.

Installation

Normally, you should install it from CRAN with

install.packages('keyToEnglish')

The latest version on GitHub can be installed via

install.packages('devtools')
devtools::install_github("mcandocia/keyToEnglish")

Note: this should ideally be done from a new R session.

Usage

The simplest way to use the package is by using the default word list, wl_common, and having fields as follows:

keyToEnglish(1:5)
# [1] "ProcedureCombAdmitCountryVoice"           "ParentPericarpNotionPompousTreat"        
# [3] "BuckSlackenReflectPublicationDeaden"      "AssociatedAldehydePastDisgraceOppressive"
# [5] "CrimsonHeelParasiteWritBenefit"  

You can also provide your own word list to the word_list parameter of keyToEnglish().

If you provide a list of lists, you can create a phrase or sentence with a specific structure. wml_long_sentence is provided, which can create sentences such as

Why would anyone need this? The main reason I could think of is being able to remember a string or other value that has been anonymized while reducing the chance of a collision.

There are also additional functions for building word lists in the package, but I haven’t documented/exported most of them, apart from corpora_to_word_list(), which takes a list of files and builds a word list by reading them. It can also parse JSON, too, which can be useful for API-retrieved data.

# example loading files from a directory
my_word_list = corpora_to_word_list(
  dir('/some/directory/with/text/files'),
  max_size=16^4
)

# example loading JSON files downloaded from Wikipedia
my_word_list = corpora_to_word_list(
  dir('/some/directory/with/json/files'),
  max_size=16^4,
  json_path=c('query','export','*')
)

By default, there are 5 phrases with 16^3 combinations, which is about 1.15 * 10^18. This should be somewhat safe up until 100 million values, where it becomes more likely you will see collisions. If you increase it to 8, then it is safe to about 13 trillion values.

Future

• Improved text-cleaning capabilities
• Even more word lists
• More options for creating phrase hashes from keyToEnglish()
• Expose some common data-downloading functions (e.g., Project Gutenberg, Wikipedia API)

Author

Max Candocia

License

GPL (>= 2)

Word lists from Michael Wehar’s repository are under public domain.