147 lines
6.7 KiB
Markdown
147 lines
6.7 KiB
Markdown
[](https://travis-ci.org/matthieugomez/StringDistances.jl)
|
|
[](https://coveralls.io/r/matthieugomez/StringDistances.jl?branch=master)
|
|
[](http://pkg.julialang.org/?pkg=StringDistances)
|
|
|
|
This Julia package computes various distances between strings.
|
|
|
|
|
|
|
|
## Distances
|
|
|
|
#### Edit Distances
|
|
- [Hamming Distance](https://en.wikipedia.org/wiki/Hamming_distance)
|
|
- [Levenshtein Distance](https://en.wikipedia.org/wiki/Levenshtein_distance)
|
|
- [Damerau-Levenshtein Distance](https://en.wikipedia.org/wiki/Damerau%E2%80%93Levenshtein_distance)
|
|
|
|
#### Q-Grams Distances
|
|
Q-gram distances compare the set of all substrings of length `q` in each string.
|
|
- QGram Distance
|
|
- [Cosine Distance](https://en.wikipedia.org/wiki/Cosine_similarity)
|
|
- [Jaccard Distance](https://en.wikipedia.org/wiki/Jaccard_index)
|
|
- [Overlap Distance](https://en.wikipedia.org/wiki/Overlap_coefficient)
|
|
- [Sorensen-Dice Distance](https://en.wikipedia.org/wiki/S%C3%B8rensen%E2%80%93Dice_coefficient)
|
|
|
|
#### Others
|
|
- [Jaro Distance](https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance)
|
|
- [RatcliffObershelp Distance](https://xlinux.nist.gov/dads/HTML/ratcliffObershelp.html)
|
|
|
|
## Syntax
|
|
#### evaluate
|
|
The function `evaluate` returns the litteral *distance* between two strings (a value of 0 being identical). While some distances are bounded by 1, other distances like `Hamming`, `Levenshtein`, `Damerau-Levenshtein`, `Jaccard` can be higher than 1.
|
|
```julia
|
|
using StringDistances
|
|
evaluate(Hamming(), "martha", "marhta")
|
|
#> 2
|
|
evaluate(QGram(2), "martha", "marhta")
|
|
#> 6
|
|
```
|
|
|
|
#### compare
|
|
The higher level function `compare` returns *a similarity score* between two strings, based on the inverse of the distance between two strings. The similarity score is always between 0 and 1. A value of 0 being completely different and a value of 1 being completely similar.
|
|
```julia
|
|
using StringDistances
|
|
compare(Hamming(), "martha", "marhta")
|
|
#> 0.6666666666666667
|
|
compare(QGram(2), "martha", "marhta")
|
|
#> 0.4
|
|
```
|
|
|
|
|
|
## Modifiers
|
|
|
|
The package defines a number of ways to modify string metrics:
|
|
|
|
- [Winkler](https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance) boosts the similary score of strings with common prefixes
|
|
|
|
```julia
|
|
compare(Jaro(), "martha", "marhta")
|
|
#> 0.9444444444444445
|
|
compare(Winkler(Jaro()), "martha", "marhta")
|
|
#> 0.9611111111111111
|
|
```
|
|
The Winkler adjustment was originally defined for the Jaro distance but this package defines it for any string distance.
|
|
|
|
```julia
|
|
compare(QGram(2), "william", "williams")
|
|
#> 0.9230769230769231
|
|
compare(Winkler(QGram(2)), "william", "williams")
|
|
#> 0.9538461538461539
|
|
```
|
|
|
|
- The Python library [fuzzywuzzy](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/) defines a few modifiers for the `RatcliffObershelp` distance. This package replicates them and extends them to any string distance:
|
|
|
|
- [Partial](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/) returns the maximal similarity score between the shorter string and substrings of the longer string.
|
|
|
|
```julia
|
|
compare(Levenshtein(), "New York Yankees", "Yankees")
|
|
#> 0.4375
|
|
compare(Partial(Levenshtein()), "New York Yankees", "Yankees")
|
|
#> 1.0
|
|
```
|
|
|
|
- [TokenSort](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/) adjusts for differences in word orders by reording words alphabetically.
|
|
|
|
```julia
|
|
compare(RatcliffObershelp(), "mariners vs angels", "angels vs mariners")
|
|
#> 0.44444
|
|
compare(TokenSort(RatcliffObershelp()),"mariners vs angels", "angels vs mariners")
|
|
#> 1.0
|
|
```
|
|
|
|
- [TokenSet](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/) adjusts for differences in word orders and word numbers by comparing the intersection of two strings with each string.
|
|
|
|
```julia
|
|
compare(Jaro(),"mariners vs angels", "los angeles angels at seattle mariners")
|
|
#> 0.559904
|
|
compare(TokenSet(Jaro()),"mariners vs angels", "los angeles angels at seattle mariners")
|
|
#> 0.944444
|
|
```
|
|
|
|
|
|
- [TokenMax](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/) combines scores using the base distance, the `Partial`, `TokenSort` and `TokenSet` modifiers, with penalty terms depending on string lengths.
|
|
```julia
|
|
compare(TokenMax(RatcliffObershelp()),"mariners vs angels", "los angeles angels at seattle mariners")
|
|
#> 0.855
|
|
```
|
|
|
|
|
|
|
|
- You can compose multiple modifiers:
|
|
```julia
|
|
compare(Winkler(Partial(Jaro())),"mariners vs angels", "los angeles angels at seattle mariners")
|
|
#> 0.7378917378917379
|
|
compare(TokenSet(Partial(RatcliffObershel())),"mariners vs angels", "los angeles angels at seattle mariners")
|
|
#> 1.0
|
|
```
|
|
|
|
|
|
## Tips
|
|
|
|
- Each distance is tailored to a specific problem. Edit distances works well with local spelling errors, the Ratcliff-Obsershelp distance works well with edited texts, the Jaro Winkler distance was invented for short strings such as person names, the QGrams distances works well with strings composed of multiple words and fluctuating orderings.
|
|
- Most distances perform poorly when comparing company or individual names, where each string is composed of multiple words.
|
|
|
|
- While word order is mostly irrelevant in this situation, edit distances heavily penalize different orderings. Instead, either use a distance robust to word order (like QGram distances), or compose a distance with `TokenSort`, which reorders the words alphabetically.
|
|
|
|
```julia
|
|
compare(RatcliffObershelp(), "mariners vs angels", "angels vs mariners")
|
|
#> 0.44444
|
|
compare(TokenSort(RatcliffObershelp()),"mariners vs angels", "angels vs mariners")
|
|
#> 1.0
|
|
compare(Cosine(3), "mariners vs angels", "angels vs mariners")
|
|
#> 0.8125
|
|
```
|
|
- General words (like "bank", "company") may appear in one string but no the other. One solution is to abbreviate these common names to diminish their importance (ie "bank" -> "bk", "company" -> "co"). Another solution is to use the `Overlap` distance, which compares common qgrams to the length of the shorter strings. Another solution is to use the `Partial` or `TokenSet` modifiers.
|
|
|
|
`TokenMax(RatcliffObershelp())`, corresponding to the `WRatio` function in the Python library `fuzzywuzzy`, combines these two behaviors and may work best in this situation.
|
|
|
|
- Standardize strings before comparing them (lowercase, punctuation, whitespaces, accents, abbreviations...)
|
|
|
|
|
|
|
|
|
|
## References
|
|
- [The stringdist Package for Approximate String Matching](https://journal.r-project.org/archive/2014-1/loo.pdf) Mark P.J. van der Loo
|
|
- [fuzzywuzzy blog post](http://chairnerd.seatgeek.com/fuzzywuzzy-fuzzy-string-matching-in-python/)
|
|
|
|
|