Fetching contributors…
Cannot retrieve contributors at this time
259 lines (175 sloc) 6.24 KB


cologne_phonetics build cologne_phonetics coverage

PyPI version



Cologne-phonetics is a phonetic algorithm similar to Soundex, wich encodes words into a phonetic code, making it possible to compare how they sound rather than how they're written. It was developed by Hans Postel and contrary to Soundex, it's designed specific for the german language.

It involves three steps:

  • Generate a code by representing every letter from left to right with a digit, according to a conversion table
  • Remove double digits
  • Remove every occurence of '0', except as a leading digit

The module itself is quite simple and consists only of the encode and compare functions and a simple command line interface.


$ "peter pédter"
127, 127
$ "umwelt umhwält"
06352, 06352
$ "urlaub uhrlaup"
0751, 0751

As you can see, similar sounding names produce the same result, with respect to the correct pronunciation.

$ "peter peta"
127, 12

This does not give the same result for each word because they may look similar, but (when pronounced correctly) don't really sound alike.

Try it / API

You can try out this implementation or use its API.


cologne_phonetics runs with Python 3.4+ or PyPy 3.5. It is available on PyPi and can be installed it via pip:

pip install cologne_phonetics

Alternatively you can download the latest release directly.


Module contents

encode(data, concat=False)

Return a list of result tuples.

Each tuple consists of the string that was encoded and its result.

If the input string is altered in any way before encoding, the tuple will contain the altered version.

>>> cologne_phonetics.encode("bäteS")
>>> [('baetes', '128')]

If concat=True is passed, words connected with hyphens will be treated as a single words.

Most of the time, the list will be len(result_list) == 1. Only if the input string contains a space character or a hyphen it is splitted into substrings and each substring will be encoded seperately.

compare(*data, concat=False)
*data. Either at last 2 positional arguments or an iterable
True if all encoded strings are equal, else False
ValueError. If only one value is submitted or the submitted Iterable is of lenght 1.

Command line interface

$ hello
$ hello world
05, 3752

Optional arguments

-h, --help show this help message and exit
-c, --concat treat words connected by hyphens as seperate words
-v, --verbose show detailed information
-p, --pretty format output nicely

Special characters

Special characters are all characters that are not ascii-characters between A and Z. Most special characters are simply ignored, but even within the set of special characters, there are some that are even more special.

Word breaks and hyphens

By default, words connected by hyphens, e.g. meier-lüdenscheid are seperated. So meier-lüdenscheid would become '67', '52682'. If you want it to be treated as a single word, you can pass a concat=True to the encode functions.

While at first this doesn't seem to make a difference in the result, other than it being split into a list of strings, in some cases it can make a difference.

>>> cologne_phonetics.encode("weiss-chemie")
>>> [('weiss', '38'), ('chemie', '46')]
>>> cologne_phonetics.encode("weiss-chemie", concat=True)
>>> [('weiss-chemie', '386')]

As you can see, a 4 got lost here. In case you really want to compare the concatenated words you may use this option, but in general there's not much use to it.

Umlaut and special character replacement

Umlaute and some other special characters are converted to their non-special equivalent.

Umlaut conversion
ü ue
ö oe
ä ae
ß s
é e
è e
á a
à a



  • Removed encode_many()
  • encode() now allways returns a list of result tuples
  • Added --verbose and --pretty options to CLI
  • New function: compare()


  • Fixed an error that would lead to case sensitive comparison in compare


  • Another error in compare was found (and fixed); Compare didn't actually compare output. It compared input. This was due to bad tests and introduced in 1.2.0, with the change that made encode always return a tuple as a result


  • PyPy 3.5 is now officially supported
  • A bug was fixed thah would lead encode to sometimes an preprocessed rather than the altered string in the result tuple