Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner:[3221] trunk
|
From: <jen...@us...> - 2011-09-02 07:50:34
|
Revision: 3221
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=3221&view=rev
Author: jenslehmann
Date: 2011-09-02 07:50:27 +0000 (Fri, 02 Sep 2011)
Log Message:
-----------
fixed bug and improved formatting in enrichment script
Modified Paths:
--------------
trunk/components-core/src/main/java/org/dllearner/utilities/CommonPrefixMap.java
trunk/interfaces/src/main/java/org/dllearner/cli/Enrichment.java
Modified: trunk/components-core/src/main/java/org/dllearner/utilities/CommonPrefixMap.java
===================================================================
--- trunk/components-core/src/main/java/org/dllearner/utilities/CommonPrefixMap.java 2011-09-02 07:22:35 UTC (rev 3220)
+++ trunk/components-core/src/main/java/org/dllearner/utilities/CommonPrefixMap.java 2011-09-02 07:50:27 UTC (rev 3221)
@@ -24,6 +24,8 @@
/**
* Can be used as base for a prefix map.
*
+ * TODO: We might implement a version of this class, which is synchronised with prefix.cc.
+ *
* @author Jens Lehmann
*
*/
@@ -32,8 +34,11 @@
private static final long serialVersionUID = 5434065917532534702L;
public CommonPrefixMap() {
+ put("dbpedia","http://dbpedia.org/resource/");
put("dbp","http://dbpedia.org/property/");
put("dbo","http://dbpedia.org/ontology/");
+ put("yago","http://dbpedia.org/class/yago/");
+ put("gml","http://www.opengis.net/gml/");
}
}
Modified: trunk/interfaces/src/main/java/org/dllearner/cli/Enrichment.java
===================================================================
--- trunk/interfaces/src/main/java/org/dllearner/cli/Enrichment.java 2011-09-02 07:22:35 UTC (rev 3220)
+++ trunk/interfaces/src/main/java/org/dllearner/cli/Enrichment.java 2011-09-02 07:50:27 UTC (rev 3221)
@@ -77,7 +77,6 @@
import org.dllearner.core.AxiomLearningAlgorithm;
import org.dllearner.core.ComponentAnn;
import org.dllearner.core.ComponentInitException;
-import org.dllearner.core.ComponentManager;
import org.dllearner.core.EvaluatedAxiom;
import org.dllearner.core.EvaluatedDescription;
import org.dllearner.core.LearningAlgorithm;
@@ -128,7 +127,6 @@
import com.hp.hpl.jena.query.ResultSet;
import com.hp.hpl.jena.rdf.model.Model;
import com.hp.hpl.jena.rdf.model.ModelFactory;
-import com.hp.hpl.jena.rdf.model.Statement;
/**
* Command Line Interface for Enrichment.
@@ -200,6 +198,10 @@
private CommonPrefixMap prefixes = new CommonPrefixMap();
+ // cache for SparqKnowledgeSource
+ SparqlKnowledgeSource ksCached;
+ AbstractReasonerComponent rcCached;
+
public Enrichment(SparqlEndpoint se, Entity resource, double threshold, boolean verbose) {
this.se = se;
this.resource = resource;
@@ -303,8 +305,8 @@
// System.out.println("Running algorithms for class " + nc);
for (Class<? extends LearningAlgorithm> algorithmClass : classAlgorithms) {
if(algorithmClass == CELOE.class) {
- applyCELOE(ks, nc, false);
- applyCELOE(ks, nc, true);
+ applyCELOE(ks, nc, false, false);
+ applyCELOE(ks, nc, true, true);
} else {
applyLearningAlgorithm((Class<AxiomLearningAlgorithm>)algorithmClass, ks, nc);
}
@@ -325,73 +327,67 @@
}
}
- private List<EvaluatedAxiom> applyCELOE(SparqlEndpointKS ks, NamedClass nc, boolean equivalence) throws ComponentInitException {
-// SPARQLTasks st = new SPARQLTasks(se);
-
+ private List<EvaluatedAxiom> applyCELOE(SparqlEndpointKS ks, NamedClass nc, boolean equivalence, boolean reuseKnowledgeSource) throws ComponentInitException {
+
// get instances of class as positive examples
SPARQLReasoner sr = new SPARQLReasoner(ks);
SortedSet<Individual> posExamples = sr.getIndividuals(nc, 20);
SortedSet<String> posExStr = Helper.getStringSet(posExamples);
- // get negative examples via various strategies
-// AutomaticNegativeExampleFinderSPARQL finder = new AutomaticNegativeExampleFinderSPARQL(posExStr, st, null);
-// finder.makeNegativeExamplesFromNearbyClasses(posExStr, 50);
-// finder.makeNegativeExamplesFromParallelClasses(posExStr, 50);
-// finder.makeNegativeExamplesFromRelatedInstances(posExStr, "http://dbpedia.org/resource/");
-// finder.makeNegativeExamplesFromSuperClasses(resource.getName(), 50);
-// finder.makeNegativeExamplesFromRandomInstances();
-// SortedSet<String> negExStr = finder.getNegativeExamples(50, false);
// use own implementation of negative example finder
+ long startTime = System.currentTimeMillis();
System.out.print("finding negatives ... ");
AutomaticNegativeExampleFinderSPARQL2 finder = new AutomaticNegativeExampleFinderSPARQL2(ks.getEndpoint());
SortedSet<String> negExStr = finder.getNegativeExamples(nc.getName(), posExStr);
negExStr = SetManipulation.fuzzyShrink(negExStr, 20);
SortedSet<Individual> negExamples = Helper.getIndividualSet(negExStr);
SortedSetTuple<Individual> examples = new SortedSetTuple<Individual>(posExamples, negExamples);
+ long runTime = System.currentTimeMillis() - startTime;
+ System.out.println("done (" + negExStr.size()+ " examples fround in " + runTime + " ms)");
- System.out.println("done (" + negExStr.size()+ ")");
-
- SparqlKnowledgeSource ks2 = new SparqlKnowledgeSource();
- ks2.setInstances(Datastructures.individualSetToStringSet(examples.getCompleteSet()));
- ks2.setUrl(ks.getEndpoint().getURL());
- ks2.setDefaultGraphURIs(new TreeSet<String>(ks.getEndpoint().getDefaultGraphURIs()));
- ks2.setUseLits(false);
- ks2.setUseCacheDatabase(true);
- ks2.setRecursionDepth(2);
- ks2.setCloseAfterRecursion(true);
-// ks2.getConfigurator().setSaveExtractedFragment(true);
- System.out.println("getting fragment ... ");
- ks2.init();
- System.out.println("done");
+ SparqlKnowledgeSource ks2;
+ AbstractReasonerComponent rc;
+ if(reuseKnowledgeSource) {
+ ks2 = ksCached;
+ rc = rcCached;
+ System.out.println("re-using previously generated knowledge base fragment");
+ } else {
+ ks2 = new SparqlKnowledgeSource();
+ ks2.setInstances(Datastructures.individualSetToStringSet(examples.getCompleteSet()));
+ ks2.setUrl(ks.getEndpoint().getURL());
+ ks2.setDefaultGraphURIs(new TreeSet<String>(ks.getEndpoint().getDefaultGraphURIs()));
+ ks2.setUseLits(false);
+ ks2.setUseCacheDatabase(true);
+ ks2.setRecursionDepth(2);
+ ks2.setCloseAfterRecursion(true);
+ startTime = System.currentTimeMillis();
+ System.out.print("getting knowledge base fragment ... ");
+ ks2.init();
+ runTime = System.currentTimeMillis() - startTime;
+ System.out.println("done in " + runTime + " ms");
+ rc = new FastInstanceChecker(ks2);
+ rc.init();
+ ksCached = ks2;
+ rcCached = rc;
+ }
- AbstractReasonerComponent rc = new FastInstanceChecker(ks2);
- rc.init();
-
- // TODO: super class learning
ClassLearningProblem lp = new ClassLearningProblem(rc);
-// lp.setPositiveExamples(posExamples);
-// lp.setNegativeExamples(negExamples);
-// try {
- lp.setClassToDescribe(nc);
-// } catch (MalformedURLException e1) {
-// e1.printStackTrace();
-// }
-// lp.setType("equivalence");
- lp.setEquivalence(true);
-// lp.setAccuracyMethod("fmeasure");
+ lp.setClassToDescribe(nc);
+ lp.setEquivalence(equivalence);
lp.setHeuristic(HeuristicType.FMEASURE);
lp.setUseApproximations(false);
lp.setMaxExecutionTimeInSeconds(10);
lp.init();
CELOE la = new CELOE(lp, rc);
-// CELOEConfigurator cc = la.getConfigurator();
la.setMaxExecutionTimeInSeconds(10);
la.setNoisePercentage(25);
la.init();
- System.out.print("running CELOE (for " + (equivalence ? "EquivalentClasses" : "SubClasses)") + "... ");
+ startTime = System.currentTimeMillis();
+ System.out.print("running CELOE (for " + (equivalence ? "equivalent classes" : "sub classes") + ") ... ");
la.start();
- System.out.println("done");
+ runTime = System.currentTimeMillis() - startTime;
+ System.out.println("done in " + runTime + " ms");
// convert the result to axioms (to make it compatible with the other algorithms)
List<? extends EvaluatedDescription> learnedDescriptions = la.getCurrentlyBestEvaluatedDescriptions(threshold);
@@ -440,7 +436,7 @@
}
}
long runtime = System.currentTimeMillis() - startTime;
- System.out.println("done in " + runtime + "ms");
+ System.out.println("done in " + runtime + " ms");
List<EvaluatedAxiom> learnedAxioms = learner
.getCurrentlyBestEvaluatedAxioms(nrOfAxiomsToLearn, threshold);
System.out.println(prettyPrint(learnedAxioms));
@@ -464,8 +460,8 @@
private String prettyPrint(EvaluatedAxiom axiom) {
double acc = axiom.getScore().getAccuracy() * 100;
String accs = df.format(acc);
- if(acc<10d) { accs = " " + accs; }
- if(acc<100d) { accs = " " + accs; }
+ if(accs.length()==3) { accs = " " + accs; }
+ if(accs.length()==4) { accs = " " + accs; }
String str = accs + "%\t" + axiom.getAxiom().toManchesterSyntaxString(null, prefixes);
return str;
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
