langcmp is a language comparison tool (written in Python 3) which computes the Levenshtein distances between the words contained in an input file and outputs, for each considered input word, a list of words containing its "closest" words. If the contents of the input file represent the most commonly used words of a language, the results will indicate how similar the words of this language are (only written form is taken into consideration; langcmp does not deal with word pronunciation). As an example of what it can be used for, see this article.

All code from this project is licensed under the GPLv3. See the LICENSE file for more information.

The following modules are used:

• matplotlib
• numpy

You can install them with the following command:

pip3 install matplotlib numpy

langcmp can break the computing work into subprocesses to reduce the overall computation time. I recommend you try using different numbers of subprocesses to find out the optimal value for your machine (suggestion: try first using the number of CPU cores available).

After computing all necessary Levenshtein distances, langcmp generates a histogram showing the fraction of the total number of words versus the number of detected words within distance d, i.e., each column in the histogram represents how many words in the input dictionary have d "closest" neighbors in this dictionary. The more "to the left" the histogram is, the bigger is the difference in spelling between the words in the dictionary.

The example command below shows most options from langcmp. It instructs langcmp to run 3 subprocesses, to consider only words in the input file words.txt which are at least 5 characters long and to only consider pairs of words which are no farther (in distance) than 2 edits from each other:

./langcmp -v -n 3 -l 5 -d 2 -i words.txt -o results.txt -s stats.txt -g histogram.txt

Above -n (--num-subproc) specifies the number of subprocesses, -l (--min-length) specifies the minimum length a word must have to be analyzed and -d (--max-distance) specifies the maximum Levenshtein distance which will be accepted in the analysis (pairs of words whose Levenshtein distances are greater than this specified value will not be considered). The results will be written on results.txt, the computed statistics on stats.txt and the histogram data on histogram.txt.

For more details on the parameters which langcmp can take, run ./langcmp -h.

This project already comes with the following word lists (in the wordlists subdirectory):

• 100 most commonly used English/German/French/Dutch words
• 1000 most commonly used English/German/French/Dutch words
• 10000 most commonly used English/German/French/Dutch words

These lists were obtaind from the University of Leipzig, Germany.

Diego Assencio / [email protected]