org.aksw.limes.core.measures.measure.string

Class JaroWinklerMeasure

java.lang.Object

org.aksw.limes.core.measures.measure.AMeasure

org.aksw.limes.core.measures.measure.string.StringMeasure

org.aksw.limes.core.measures.measure.string.JaroWinklerMeasure

All Implemented Interfaces:

IMeasure, IStringMeasure, ITrieFilterableStringMeasure

public class JaroWinklerMeasure
extends StringMeasure
implements ITrieFilterableStringMeasure

This class implements the Jaro-Winkler algorithm that was designed as a string subsequence alignment method for matching names in the US Census. It is thus optimized for relatively small sized strings of latin letters only. It provides all the features of the original C implementation by William E. Winkler, although the features that made it specific for name matching may be disabled.

To overcome the complexity O(n*m) for non matching cases a filter is added. Given a threshold it can identify pairs whose Jaro-Winkler proximity is confidently less than or equal to that threshold.

Author:

Kevin Dreßler

Field Summary

Fields

Modifier and Type	Field and Description
static double	winklerBoostThreshold

Constructor Summary

Constructors

Constructor and Description
JaroWinklerMeasure()
JaroWinklerMeasure(boolean uppercaseOn, boolean longStringsOn, boolean characterSimilarityOn)

Method Summary

Modifier and Type	Method and Description
double	characterFrequencyUpperBound(int l1, int l2, int m)
int	characterMatchLowerBound(int l1, int l2, double threshold)
JaroWinklerMeasure	clone() Clone method for parallel execution
boolean	computableViaOverlap() Returns true if this similarity function can be computed just via the getSimilarity(overlag, lengthA, lengthB)
int	getAlpha(int xTokensNumber, int yTokensNumber, double threshold) Threshold for the positional filtering
char[]	getArrayRepresentation(String s)
int	getMidLength(int tokensNumber, double threshold) Theshold for the length of the tokens to be indexed
String	getName() Returns name of a measure.
LinkedList<org.apache.commons.lang3.tuple.ImmutableTriple<Integer,Integer,Integer>>	getPartitionBounds(int maxSize, double threshold)
int	getPrefixLength(int tokensNumber, double threshold) Length of prefix to consider when mapping the input string with other strings.
double	getRuntimeApproximation(double mappingSize) Returns the runtime approximation of a measure.
double	getSimilarity(Instance instance1, Instance instance2, String property1, String property2) Returns the similarity between two instances, given their corresponding properties.
double	getSimilarity(int overlap, int lengthA, int lengthB) Returns the similarity of two strings given their length and the overlap.
double	getSimilarity(Object object1, Object object2) Returns the similarity between two objects.
double	getSizeFilteringThreshold(int tokensNumber, double threshold)
String	getType() Returns type of a measure.
int	lengthLowerBound(int l1, double threshold)
int	lengthUpperBound(int l1, double threshold)
double	proximity(char[] yin, char[] yang) Calculate the proximity of two input strings if proximity is assured to be over given threshold threshold.
double	proximity(String yi, String ya) Calculate the proximity of two input strings if proximity is assured to be over given threshold threshold.

Methods inherited from class java.lang.Object

equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

winklerBoostThreshold

public static double winklerBoostThreshold

Constructor Detail

JaroWinklerMeasure

public JaroWinklerMeasure()

JaroWinklerMeasure

public JaroWinklerMeasure(boolean uppercaseOn,
                          boolean longStringsOn,
                          boolean characterSimilarityOn)

Method Detail

clone

public JaroWinklerMeasure clone()

Clone method for parallel execution

Overrides:

clone in class Object

Returns:

clone

proximity

public double proximity(char[] yin,
                        char[] yang)

Calculate the proximity of two input strings if proximity is assured to be over given threshold threshold.

Parameters:

yin - string to align on

yang - string to align on

Returns:

similarity score (proximity)

proximity

public double proximity(String yi,
                        String ya)

Calculate the proximity of two input strings if proximity is assured to be over given threshold threshold.

Specified by:

proximity in interface ITrieFilterableStringMeasure

Parameters:

yi - string to be aligned

ya - string to align on

Returns:

similarity score (proximity)

getArrayRepresentation

public char[] getArrayRepresentation(String s)

characterFrequencyUpperBound

public double characterFrequencyUpperBound(int l1,
                                           int l2,
                                           int m)

Specified by:

characterFrequencyUpperBound in interface ITrieFilterableStringMeasure

characterMatchLowerBound

public int characterMatchLowerBound(int l1,
                                    int l2,
                                    double threshold)

Specified by:

characterMatchLowerBound in interface ITrieFilterableStringMeasure

lengthUpperBound

public int lengthUpperBound(int l1,
                            double threshold)

Specified by:

lengthUpperBound in interface ITrieFilterableStringMeasure

lengthLowerBound

public int lengthLowerBound(int l1,
                            double threshold)

Specified by:

lengthLowerBound in interface ITrieFilterableStringMeasure

getPartitionBounds

public LinkedList<org.apache.commons.lang3.tuple.ImmutableTriple<Integer,Integer,Integer>> getPartitionBounds(int maxSize,
                                                                                                              double threshold)

Specified by:

getPartitionBounds in interface ITrieFilterableStringMeasure

getPrefixLength

public int getPrefixLength(int tokensNumber,
                           double threshold)

Description copied from interface: IStringMeasure

Length of prefix to consider when mapping the input string with other strings.

Specified by:

getPrefixLength in interface IStringMeasure

Parameters:

tokensNumber - Size of input string in

threshold - Similarity threshold

Returns:

Prefix length

getMidLength

public int getMidLength(int tokensNumber,
                        double threshold)

Description copied from interface: IStringMeasure

Theshold for the length of the tokens to be indexed

Specified by:

getMidLength in interface IStringMeasure

Parameters:

tokensNumber - Number of tokens of current input

threshold - Similarity threshold

Returns:

Length of tokens to be indexed

getSizeFilteringThreshold

public double getSizeFilteringThreshold(int tokensNumber,
                                        double threshold)

Specified by:

getSizeFilteringThreshold in interface IStringMeasure

getAlpha

public int getAlpha(int xTokensNumber,
                    int yTokensNumber,
                    double threshold)

Description copied from interface: IStringMeasure

Threshold for the positional filtering

Specified by:

getAlpha in interface IStringMeasure

Parameters:

xTokensNumber - Size of the first input string

yTokensNumber - Size of the first input string

threshold - Similarity threshold

Returns:

Threshold for positional filtering

getSimilarity

public double getSimilarity(int overlap,
                            int lengthA,
                            int lengthB)

Description copied from interface: IStringMeasure

Returns the similarity of two strings given their length and the overlap. Useful when these values are known so that no computation of known values have to be carried out anew

Specified by:

getSimilarity in interface IStringMeasure

Parameters:

overlap - Overlap of strings A and B

lengthA - Length of A

lengthB - Length of B

Returns:

Similarity of A and B

computableViaOverlap

public boolean computableViaOverlap()

Description copied from interface: IStringMeasure

Returns true if this similarity function can be computed just via the getSimilarity(overlag, lengthA, lengthB)

Specified by:

computableViaOverlap in interface IStringMeasure

Returns:

True if it's possible, else false;

getSimilarity

public double getSimilarity(Object object1,
                            Object object2)

Description copied from interface: IMeasure

Returns the similarity between two objects.

Specified by:

getSimilarity in interface IMeasure

Returns:

The similarity of the objects

getType

public String getType()

Description copied from interface: IMeasure

Returns type of a measure.

Specified by:

getType in interface IMeasure

Returns:

The runtime of the measure

getSimilarity

public double getSimilarity(Instance instance1,
                            Instance instance2,
                            String property1,
                            String property2)

Description copied from interface: IMeasure

Returns the similarity between two instances, given their corresponding properties.

Specified by:

getSimilarity in interface IMeasure

Returns:

The similarity of the instances

getName

public String getName()

Description copied from interface: IMeasure

Returns name of a measure.

Specified by:

getName in interface IMeasure

Returns:

Measure name as a string

getRuntimeApproximation

public double getRuntimeApproximation(double mappingSize)

Description copied from interface: IMeasure

Returns the runtime approximation of a measure.

Specified by:

getRuntimeApproximation in interface IMeasure

Returns:

The runtime of the measure