org.dllearner.scripts.improveWikipedia
Class WikipediaCategoryTasks

java.lang.Object
  org.dllearner.scripts.improveWikipedia.WikipediaCategoryTasks

public class WikipediaCategoryTasks
extends Object

Constructor Summary

WikipediaCategoryTasks(SPARQLTasks sparqlTasks)

Method Summary

SortedSet<String>	calculateWrongIndividualsAndNewPosEx(List<EvaluatedDescriptionPosNeg> conceptresults, SortedSet<String> posExamples) The strategy is yet really simple.
SortedSet<String>	getCleanedPositiveSet()
SortedSet<String>	getDefinitelyWrongIndividuals()
SortedSet<String>	getFullPositiveSet()
SortedSet<String>	getNegExamples()
SortedSet<String>	getPosExamples()
void	makeInitialExamples(String targetCategory, double percentOfSKOSSet, double negFactor, int sparqlResultLimitNegativeExamples, boolean stable) makes positive and negative Examples. positives are a simple retrieval of the category. negatives are made from parallelclasses.
SortedSet<String>	makeNewNegativeExamples(List<EvaluatedDescriptionPosNeg> reEvaluatedDesc, SortedSet<String> posExamples, double negFactor) TODO could be more sophisticated

Methods inherited from class java.lang.Object

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

Constructor Detail

WikipediaCategoryTasks

public WikipediaCategoryTasks(SPARQLTasks sparqlTasks)

Method Detail

calculateWrongIndividualsAndNewPosEx

public SortedSet<String> calculateWrongIndividualsAndNewPosEx(List<EvaluatedDescriptionPosNeg> conceptresults,
                                                              SortedSet<String> posExamples)

The strategy is yet really simple. //TODO take the best concept and the notCoveredPositives are the ones definitely wrong these are removed from the positives examples.

Parameters:

conceptresults -

posExamples -

makeNewNegativeExamples

public SortedSet<String> makeNewNegativeExamples(List<EvaluatedDescriptionPosNeg> reEvaluatedDesc,
                                                 SortedSet<String> posExamples,
                                                 double negFactor)

TODO could be more sophisticated

Parameters:

reEvaluatedDesc -

makeInitialExamples

public void makeInitialExamples(String targetCategory,
                                double percentOfSKOSSet,
                                double negFactor,
                                int sparqlResultLimitNegativeExamples,
                                boolean stable)

makes positive and negative Examples. positives are a simple retrieval of the category. negatives are made from parallelclasses.

Parameters:

targetCategory -

percentOfSKOSSet - percentage used from the SKOSSet for training

negFactor - size of the negative Examples compared to the posExample size (1.0 means equal size)

sparqlResultLimit -

getPosExamples

public SortedSet<String> getPosExamples()

getNegExamples

public SortedSet<String> getNegExamples()

getFullPositiveSet

public SortedSet<String> getFullPositiveSet()

getDefinitelyWrongIndividuals

public SortedSet<String> getDefinitelyWrongIndividuals()

getCleanedPositiveSet

public SortedSet<String> getCleanedPositiveSet()