Package com.wcohen.ss

This package contains a bunch of approximate string comparators, plus code for performing controlled experiments with this.

See:
Description

Class Summary

AbstractStatisticalTokenDistance	Abstract token distance metric that uses frequency statistics.
AbstractStringDistance	Abstract class which implements StringDistanceLearner as well as StringDistance.
AbstractTokenizedStringDistance	Abstract distance metric for tokenized strings.
AdaptiveStringDistanceLearner	Abstract StringDistanceLearner class which averages results of a number of inner distance metrics, learned by a number of inner distance learners.
AffineGap	Affine-gap string distance, following Durban et al.
ApproxMemoMatrix	Variant of MemoMatrix that only stores values near the diagonal, for better efficiency.
ApproxNeedlemanWunsch	Needleman-Wunsch string distance, following Durban et al.
AveragedStringDistanceLearner	Abstract StringDistanceLearner class which averages results of a number of inner distance metrics, learned by a number of inner distance learners.
BasicDistanceInstanceIterator	A simple DistanceInstanceIterator implementation.
BasicStringWrapper	An extendible (non-final) class that implements some of the functionality of a string.
BasicStringWrapperIterator	A simple StringWrapperIterator implementation.
CharMatchScore	Abstract distance between characters.
CombinedStringDistanceLearner	Abstract StringDistanceLearner class which combines results of a number of inner distance metrics, learned by a number of inner distance learners.
DirichletJS	Jensen-Shannon distance of two unigram language models, smoothed using Dirichlet prior.
DistanceLearnerFactory	Creates distance metric learners from string descriptions.
Jaccard	Jaccard distance implementation.
Jaro	Jaro distance metric.
JaroWinkler	Jaro distance metric, as extended by Winkler.
JaroWinklerTFIDF	Soft TFIDF-based distance metric, extended to use "soft" token-matching with the JaroWinkler distance metric.
JelinekMercerJS	Jensen-Shannon distance of two unigram language models, smoothed using Jelinek-Mercer mixture model.
JensenShannonDistance	Distance metrics based on Jensen-Shannon distance of two smoothed unigram language models.
Level2	Generic version of Monge & Elkan's "level 2" recursive field matching.
Level2Jaro	"Level 2" recursive field matching algorithm, based on Jaro distance.
Level2JaroWinkler	"Level 2" recursive field matching algorithm, based on Jaro distance.
Level2Levenstein	"Level 2" recursive field matching algorithm using Levenstein distance.
Level2MongeElkan	Monge & Elkan's "level 2" recursive field matching algorithm.
Levenstein	Levenstein string distance.
MemoMatrix	A matrix of doubles, defined recursively by the compute(i,j) method, that will not be recomputed more than necessary.
Mixture	Mixture-based distance metric.
MongeElkan	The match method proposed by Monge and Elkan.
MultiStringAvgDistance	StringDistance defined over Strings that are broken into fields, with distance defined as the average distance between any field.
MultiStringDistance	Abstract class StringDistance defined over Strings that are broken into fields.
MultiStringWrapper	A StringWrapper that stores a version of the string that has been either (a) split into a number of distinct fields, or (b) duplicated k times, so that k different StringDistance's can preprocess it, of (b) both of the above.
NeedlemanWunsch	Needleman-Wunsch string distance, following Durban et al.
PrintfFormat	PrintfFormat allows the formatting of an array of objects embedded within a string.
ScaledLevenstein	Levenstein string distance.
SmithWaterman	Smith-Waterman string distance, following Durban et al.
SoftTFIDF	TFIDF-based distance metric, extended to use "soft" token-matching.
SoftTokenFelligiSunter	Highly simplified model of Felligi-Sunter's method 1, applied to tokens.
TagLink
TagLink.Candidates
TFIDF	TFIDF-based distance metric.
TokenFelligiSunter	Highly simplified model of Felligi-Sunter's method 1, applied to tokens.
UnsmoothedJS	Jensen-Shannon distance of two unsmoothed unigram language models.
WinklerRescorer	Winkler's reweighting scheme for distance metrics.
WizardUI	Top-level GUI interface.

Package com.wcohen.ss Description

This package contains a bunch of approximate string comparators, plus code for performing controlled experiments with this.

A StringDistance is the basic class for computing distances. The score() function of this class outputs a distance measure between its two arguments. The other methods are there for efficiency, so that preprocessing steps (like tokenization) can be amortized over multiple comparisons with the same string.

The way that preprocessing steps are saved is by creating a StringWrapper object which contains the preprocessed string, plus whatever else needs to be cached. To do this, extend default implementation of StringWrapper.

Almost everything in this package implements StringDistance. The only (public) exceptions are StringWrapper; PrintfFormat, pilfered from Sun to make the explanations easier; CharMatchScore, which is a character-based distance metric; and MemoMatrix, a utility for defining edit-distance-based methods.