String Similarity and Distance Measures

0.2.1 · active · verified Wed Apr 15

Strsimpy is a Python library that provides implementations for various string similarity and distance measures, including popular algorithms like Levenshtein, Jaro-Winkler, N-Gram, Cosine Similarity, and Jaccard Index. It's designed to be straightforward to use for text analysis and data matching tasks. The current version is 0.2.1. Releases are infrequent, typically addressing bug fixes or adding new algorithms.

Warnings

breaking The package name was changed from `similarity` to `strsimpy`. Old import paths will fail.
Fix: Update all `import similarity...` statements to `import strsimpy...` and reinstall the package.
breaking The `WeightedLevenshtein` algorithm's API changed significantly in v0.1.7. It now uses functions for weight calculation instead of expecting Class or Objects, simplifying its usage.
Fix: Review the documentation for `WeightedLevenshtein` and update your weight calculation logic to use direct functions as arguments, rather than class instances.
gotcha The `numpy` dependency was removed in v0.1.5 to lighten the package. If you were implicitly relying on `numpy` being installed alongside `strsimpy`, it might no longer be present.
Fix: If your project requires `numpy`, ensure it is explicitly listed in your project's dependencies.
gotcha A `ZeroDivisionError` was fixed in ShingleBased algorithms (e.g., Jaccard, Cosine Similarity) for certain edge cases.
Fix: Upgrade to version 0.1.2 or later to ensure correct calculations and avoid potential runtime errors with ShingleBased algorithms.

Install

pip install strsimpy Install latest version

Imports

Levenshtein

from strsimpy.levenshtein import Levenshtein

JaroWinkler

from strsimpy.jaro_winkler import JaroWinkler

Ngram
```
from strsimpy.ngram import Ngram
```

WeightedLevenshtein

wrong:

from strsimpy.weighted_levenshtein import Levenshtein

correct:

from strsimpy.weighted_levenshtein import WeightedLevenshtein

Quickstart

This quickstart demonstrates calculating Levenshtein distance and Jaro-Winkler similarity between strings. Most algorithms follow a pattern of instantiating a class and then calling a `distance()` or `similarity()` method.

from strsimpy.levenshtein import Levenshtein

s0 = "안녕하세요"
s1 = "안녕하세유"

levenshtein = Levenshtein()
distance = levenshtein.distance(s0, s1)
print(f"Levenshtein distance between '{s0}' and '{s1}': {distance}")

s2 = "apple"
s3 = "aple"
distance2 = levenshtein.distance(s2, s3)
print(f"Levenshtein distance between '{s2}' and '{s3}': {distance2}")

from strsimpy.jaro_winkler import JaroWinkler

jaro_winkler = JaroWinkler()
similarity = jaro_winkler.similarity(s2, s3)
print(f"Jaro-Winkler similarity between '{s2}' and '{s3}': {similarity}")

view raw JSON →