Stringmetric

#stringmetric String metrics and phonetic algorithms for Scala. The library provides facilities to perform approximate string matching, measurement of string similarity/distance, indexing by word pronunciation, and sounds-like comparisons. In addition to the core library, each metric and algorithm has a command line interface.

• Requirements: Scala 2.10+
• Documentation: Scaladoc
• Issues: Enhancements, Questions, Bugs
• Versioning: Semantic Versioning v2.0

Metrics and algorithms

• Dice / Sorensen (Similarity metric)
• Double Metaphone (Queued phonetic metric and algorithm)
• Hamming (Similarity metric)
• Jaccard (Similarity metric)
• Jaro (Similarity metric)
• Jaro-Winkler (Similarity metric)
• Levenshtein (Similarity metric)
• Metaphone (Phonetic metric and algorithm)
• Monge-Elkan (Queued similarity metric)
• Match Rating Approach (Queued phonetic metric and algorithm)
• Needleman-Wunch (Queued similarity metric)
• N-Gram (Similarity metric)
• NYSIIS (Phonetic metric and algorithm)
• Overlap (Similarity metric)
• Ratcliff-Obershelp (Similarity metric)
• Refined NYSIIS (Phonetic metric and algorithm)
• Refined Soundex (Phonetic metric and algorithm)
• Tanimoto (Queued similarity metric)
• Tversky (Queued similarity metric)
• Smith-Waterman (Queued similarity metric)
• Soundex (Phonetic metric and algorithm)
• Weighted Levenshtein (Similarity metric)

Depending upon

SBT:

libraryDependencies += "com.rockymadden.stringmetric" %% "stringmetric-core" % "0.27.4"

---

Gradle:

compile 'com.rockymadden.stringmetric:stringmetric-core_2.10:0.27.4'

---

Maven:

<dependency>
	<groupId>com.rockymadden.stringmetric</groupId>
	<artifactId>stringmetric-core_2.10</artifactId>
	<version>0.27.4</version>
</dependency>

---

Similarity package

Useful for approximate string matching and measurement of string distance. Most metrics calculate the similarity of two strings as a double with a value between 0 and 1. A value of 0 being completely different and a value of 1 being completely similar.

---

Dice / Sorensen Metric:

DiceSorensenMetric(1).compare("night", "nacht") // 0.6
DiceSorensenMetric(1).compare("context", "contact") // 0.7142857142857143

^{Note you must specify the size of the n-gram you wish to use.}

---

Hamming Metric:

HammingMetric.compare("toned", "roses") // 3
HammingMetric.compare("1011101", "1001001") // 2

^{Note the exception of integers, rather than doubles, being returned.}

---

Jaccard Metric:

JaccardMetric(1).compare("night", "nacht") // 0.3
JaccardMetric(1).compare("context", "contact") // 0.35714285714285715

^{Note you must specify the size of the n-gram you wish to use.}

---

Jaro Metric:

JaroMetric.compare("dwayne", "duane") // 0.8222222222222223
JaroMetric.compare("jones", "johnson") // 0.7904761904761904
JaroMetric.compare("fvie", "ten") // 0.0

---

Jaro-Winkler Metric:

JaroWinklerMetric.compare("dwayne", "duane") // 0.8400000000000001
JaroWinklerMetric.compare("jones", "johnson") // 0.8323809523809523
JaroWinklerMetric.compare("fvie", "ten") // 0.0

---

Levenshtein Metric:

LevenshteinMetric.compare("sitting", "kitten") // 3
LevenshteinMetric.compare("cake", "drake") // 2

^{Note the exception of integers, rather than doubles, being returned.}

---

N-Gram Metric:

NGramMetric(1).compare("night", "nacht") // 0.6
NGramMetric(2).compare("night", "nacht") // 0.25
NGramMetric(2).compare("context", "contact") // 0.5

^{Note you must specify the size of the n-gram you wish to use.}

---

Overlap Metric:

OverlapMetric(1).compare("night", "nacht") // 0.6
OverlapMetric(1).compare("context", "contact") // 0.7142857142857143

^{Note you must specify the size of the n-gram you wish to use.}

---

Ratcliff/Obershelp Metric:

RatcliffObershelpMetric.compare("aleksander", "alexandre") // 0.7368421052631579
RatcliffObershelpMetric.compare("pennsylvania", "pencilvaneya") // 0.6666666666666666

---

Weighted Levenshtein Metric:

WeightedLevenshteinMetric(10, 0.1, 1).compare("book", "back") // 2
WeightedLevenshteinMetric(10, 0.1, 1).compare("hosp", "hospital") // 0.4
WeightedLevenshteinMetric(10, 0.1, 1).compare("hospital", "hosp") // 40

^{Note you must specify the weight of each operation. Delete, insert, and then substitute. Note that while a double is returned, it can be outside the range of 0 to 1, based upon the weights used.}

---

Phonetic package

Useful for indexing by word pronunciation and performing sounds-like comparisons. All metrics return a boolean value indicating if the two strings sound the same, per the algorithm used. All metrics have an algorithm counterpart which provide the means to perform indexing by word pronunciation.

---

Metaphone Metric:

MetaphoneMetric.compare("merci", "mercy") // true
MetaphoneMetric.compare("dumb", "gum") // false

---

Metaphone Algorithm:

MetaphoneAlgorithm.compute("dumb") // tm
MetaphoneAlgorithm.compute("knuth") // n0

---

NYSIIS Metric:

NysiisMetric.compare("ham", "hum") // true
NysiisMetric.compare("dumb", "gum") // false

---

NYSIIS Algorithm:

NysiisAlgorithm.compute("macintosh") // mcant
NysiisAlgorithm.compute("knuth") // nnat

---

Refined NYSIIS Metric:

RefinedNysiisMetric.compare("ham", "hum") // true
RefinedNysiisMetric.compare("dumb", "gum") // false

---

Refined NYSIIS Algorithm:

RefinedNysiisAlgorithm.compute("macintosh") // mcantas
RefinedNysiisAlgorithm.compute("westerlund") // wastarlad

---

Refined Soundex Metric:

RefinedSoundexMetric.compare("robert", "rupert") // true
RefinedSoundexMetric.compare("robert", "rubin") // false

---

Refined Soundex Algorithm:

RefinedSoundexAlgorithm.compute("hairs") // h093
RefinedSoundexAlgorithm.compute("lambert") // l7081096

---

Soundex Metric:

SoundexMetric.compare("robert", "rupert") // true
SoundexMetric.compare("robert", "rubin") // false

---

Soundex Algorithm:

SoundexAlgorithm.compute("rupert") // r163
SoundexAlgorithm.compute("lukasiewicz") // l222

---

Convenience objects

StringAlgorithm:

StringAlgorithm.computeWithMetaphone("abcdef")
StringAlgorithm.computeWithNysiis("abcdef")

---

StringMetric:

StringMetric.compareWithJaccard(1)("abcdef", "abcxyz")
StringMetric.compareWithJaroWinkler("abcdef", "abcxyz")

---

Decorating

It is possible to decorate algorithms and metrics with additional functionality, which you can mix and match. Decorations include:

• withMemoization: Computations and comparisons are cached. Future calls made with identical arguments will be looked up, rather than computed.

• withTransform: Transform arguments prior to computation/comparison. A handful of pre-built transforms are located in the transform module.

---

Non-decorated:

MetaphoneAlgorithm.compute("abcdef")
MetaphoneMetric.compare("abcdef", "abcxyz")

---

Using memoization:

(MetaphoneAlgorithm withMemoization).compute("abcdef")

---

Using a transform so that we only examine alphabetical characters:

(MetaphoneAlgorithm withTransform filterAlpha).compute("abcdef")
(MetaphoneMetric withTransform filterAlpha).compare("abcdef", "abcxyz")

---

Using a functionally composed transform so that we only examine alphabetical characters, but the case will not matter:

val composedTransform = (filterAlpha andThen ig