Documentation for Text

The Daitch-Mokotoff Soundex algorithm is a variant of the Russell Soundex algorithm, designed to work better for Slavic and Yiddish names. The implementation here uses the table from http://www.jewishgen.org/InfoFiles/soundex.html.

For some words, multiple codes are possible - pass an optional second argument 'all to get a list of codes:

The hamming-distance is the number of mismatched items between two equal-length sequences. The hamming-distance function takes two arguments and an optional comparison procedure. The sequences can be strings, lists, vectors or bytevectors. The comparison procedure defaults to char=? for strings, = for bytevectors, and equal? for everything else.

The levenshtein-distance counts the minimum number of insertions/deletions/substitutions to convert one sequence into another. The levenshtein-distance function takes two arguments and an optional comparison procedure. The sequences can be strings, lists, vectors or bytevectors. The comparison procedure defaults to char=? for strings, = for bytevectors, and equal? for everything else.

The Metaphone Algorithm was created as an improvement on Soundex, better taking account of variations in English pronounciation. The algorithm was created by Lawrence Philips, in "Computer Language" December 1990 issue. There have been many published variants of Metaphone. A summary can be found at http://aspell.net/metaphone/.

The metaphone function simply takes a word to convert into a coded representation of its sound:

Note, the output code is all upper-case letters, with "0" standing in for "TH".

optimal-string-alignment-distance is a modification of the Levenshtein distance to include transpositions as well as deletions, insertions or substitutions. A transposition is where two characters have been swapped, such as when typing too quiclky. The optimal-string-alignment-distance function takes two arguments and an optional comparison procedure. The sequences can be strings, lists, vectors or bytevectors. The comparison procedure defaults to char=? for strings, = for bytevectors, and equal? for everything else.

Notice the difference between the two algorithms in the last case, where "n" and "i" have been transposed.

porter-stem is an implementation of the well-known Porter Stemming Algorithm, for reducing words to a base form. More details of the algorithm are at https://tartarus.org/martin/PorterStemmer/

The function simply takes the word to change, and returns the stemmed form:

russell-soundex is the same as soundex, exported from (slib soundex)

sorenson-dice-similarity returns a measure of how similar two strings are, based on n-grams of characters. An optional third argument provides the number of characters in the n-grams, which defaults to 2:

soundex is exported from (slib soundex).

string→n-grams separates a string into overlapping groups of n letters.

For n greater than the length of the string, the string itself is returned. If n is less than 1, an error is raised.

words→with-commas is a function taking a list of strings and adding commas in between the items up to the last item, which is preceded by an "and". For example:

An optional second argument controls whether the final "and" should be preceded by a comma. The default is not to have the comma.

word-wrap takes two arguments, a string and a target width. It returns a list of strings, each item in the list representing a line of text. Each line is formatted so words do not go beyond the target width. The algorithm is a simple greedy algorithm. If a single word is longer than the target width, it is allowed to overlap.

For example, setting:

We can wrap to a width of 50 characters using:

and we can wrap to a width of 60 characters using: