string similarity in python

[Achim Domma]

Hi,

I have a list of lets say 100-1000 strings and want to know which one is
most similar to a reference string. Does somebody know such a library for
Python? I don't need complicated scientific stuff, I think the most simple
ones will do it for my data.

regards,
Achim

 

[Peter Otten]

I have a list of lets say 100-1000 strings and want to know which one is
most similar to a reference string. Does somebody know such a library for
Python? I don't need complicated scientific stuff, I think the most simple
ones will do it for my data.

Remembering an algorithm called Levenshtein, google came up with something
that looks promising.

http://trific.ath.cx/resources/python/levenshtein/

HTH,
Peter

 

[Rene Pijlman]

Achim Domma:

I have a list of strings and want to know which one is
most similar to a reference string.

Define 's1 is most similar to s2' for all strings s1 and s2.

You can skip the definition of 'most', when your definition of 'similar'
defines a total order
http://en2.wikipedia.org/wiki/Total_order

 

[anton muhin]

Hi,

I have a list of lets say 100-1000 strings and want to know which one is
most similar to a reference string. Does somebody know such a library for
Python? I don't need complicated scientific stuff, I think the most simple
ones will do it for my data.

regards,
Achim

get_close_matches function in standard difflib module might be what you
are looking for.

regards,
anton.

 

[vincent wehren]

| Hi,
|
| I have a list of lets say 100-1000 strings and want to know which one is
| most similar to a reference string. Does somebody know such a library for
| Python? I don't need complicated scientific stuff, I think the most simple
| ones will do it for my data.
|
| regards,
| Achim
|
|

http://trific.ath.cx/resources/python/levenshtein/

It lets you calculate Levenshtein distance as well as a ratio of similarity
based on it, allowing you to "tweak" your results. You can use the source
both as C app or as C/Python extension module.

Getting it to do what you probably won't take you more than a few minutes...

Regards

Vincent Wehren

 

[Luca Montecchiani]

Peter Otten wrote:
:

Remembering an algorithm called Levenshtein, google came up with something
that looks promising.

http://trific.ath.cx/resources/python/levenshtein/

:

Other resources :

o http://www-users.cs.umn.edu/~ialbert/python/apse/

o http://datamining.anu.edu.au/projects/linkage.html
( stringcmp.py encode.py modules )

o http://spazioinwind.libero.it/montecchiani/pyfdupes-0.7.tgz

o http://www.bio.cam.ac.uk/~mw263/pyagrep.html

ciao,
luca