Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner:[2034] trunk/src/dl-learner/org/dllearner
|
From: <ku...@us...> - 2010-02-13 19:33:57
|
Revision: 2034
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=2034&view=rev
Author: kurzum
Date: 2010-02-13 19:33:50 +0000 (Sat, 13 Feb 2010)
Log Message:
-----------
finished script for iterative learning
Modified Paths:
--------------
trunk/src/dl-learner/org/dllearner/examples/KRKOntologyTBox.java
trunk/src/dl-learner/org/dllearner/kb/sparql/SparqlQueryDescriptionConvertVisitor.java
trunk/src/dl-learner/org/dllearner/scripts/tiger/ExperimentConfig.java
trunk/src/dl-learner/org/dllearner/scripts/tiger/TestIterativeLearning.java
trunk/src/dl-learner/org/dllearner/utilities/examples/Examples.java
Added Paths:
-----------
trunk/src/dl-learner/org/dllearner/scripts/tiger/GlobalTest.java
Modified: trunk/src/dl-learner/org/dllearner/examples/KRKOntologyTBox.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/examples/KRKOntologyTBox.java 2010-02-13 17:45:31 UTC (rev 2033)
+++ trunk/src/dl-learner/org/dllearner/examples/KRKOntologyTBox.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -70,11 +70,13 @@
public void addConcept(String concept){
try{
//make Description
- Description d = KBParser.parseConcept(concept,ontologyURI.toString()+"#");
- //d.addChild(oecr);
+ KBParser.internalNamespace = ontologyURI.toString()+"#";
+ Description d = KBParser.parseConcept(concept);
kb.addTBoxAxiom(new EquivalentClassesAxiom(getAtomicConcept("test"),d));
- }catch (Exception e) {e.printStackTrace();}
+ }catch (Exception e) {
+ e.printStackTrace();
+ }
}
public void initOntologyTBox(){
Modified: trunk/src/dl-learner/org/dllearner/kb/sparql/SparqlQueryDescriptionConvertVisitor.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/kb/sparql/SparqlQueryDescriptionConvertVisitor.java 2010-02-13 17:45:31 UTC (rev 2033)
+++ trunk/src/dl-learner/org/dllearner/kb/sparql/SparqlQueryDescriptionConvertVisitor.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -30,9 +30,12 @@
import org.apache.log4j.Logger;
import org.dllearner.algorithms.gp.ADC;
import org.dllearner.core.ComponentManager;
+import org.dllearner.core.owl.Constant;
+import org.dllearner.core.owl.Datatype;
import org.dllearner.core.owl.DatatypeExactCardinalityRestriction;
import org.dllearner.core.owl.DatatypeMaxCardinalityRestriction;
import org.dllearner.core.owl.DatatypeMinCardinalityRestriction;
+import org.dllearner.core.owl.DatatypeProperty;
import org.dllearner.core.owl.DatatypeSomeRestriction;
import org.dllearner.core.owl.DatatypeValueRestriction;
import org.dllearner.core.owl.Description;
@@ -49,7 +52,9 @@
import org.dllearner.core.owl.ObjectProperty;
import org.dllearner.core.owl.ObjectSomeRestriction;
import org.dllearner.core.owl.ObjectValueRestriction;
+import org.dllearner.core.owl.StringValueRestriction;
import org.dllearner.core.owl.Thing;
+import org.dllearner.core.owl.TypedConstant;
import org.dllearner.core.owl.Union;
import org.dllearner.parser.KBParser;
import org.dllearner.parser.ParseException;
@@ -229,15 +234,43 @@
s.clear();
// s.add("(\"http://nlp2rdf.org/ontology/Sentence\" AND (EXISTS \"http://nlp2rdf.org/ontology/syntaxTreeHasPart\".\"http://nachhalt.sfb632.uni-potsdam.de/owl/stts.owl#Pronoun\" AND EXISTS \"http://nlp2rdf.org/ontology/syntaxTreeHasPart\".\"http://nlp2rdf.org/ontology/sentencefinalpunctuation_tag\"))");
- s.add("(\"http://nlp2rdf.org/ontology/Sentence\" AND (\"http://nlp2rdf.org/ontology/hasLemma\" VALUE \"test\" )");
+// s.add("(\"http://nlp2rdf.org/ontology/Sentence\" AND (\"http://nlp2rdf.org/ontology/hasLemma\" VALUE \"test\" )");
+ String prefix = "http://nlp2rdf.org/ontology/";
+ String test = "(\"Sentence\" AND (EXISTS \"syntaxTreeHasPart\".\"VVPP\" AND EXISTS \"syntaxTreeHasPart\".(\"stts:AuxilliaryVerb\" AND \"hasLemma\" = werden)))";
+
+ ObjectProperty stp = new ObjectProperty(prefix+"syntaxTreeHasPart");
+ DatatypeProperty dtp = new DatatypeProperty(prefix+"hasLemma");
+ StringValueRestriction svr = new StringValueRestriction(dtp,"werden" );
+ Intersection inner = new Intersection(new NamedClass(prefix+"Auxillary"), svr);
+ Intersection middle = new Intersection(
+ new ObjectSomeRestriction(stp, new NamedClass(prefix+"VVPP")),
+ new ObjectSomeRestriction(stp, inner));
+ Intersection outer = new Intersection(
+ new NamedClass(prefix+"Sentence"),
+ middle
+ );
+
+ System.out.println(outer.toKBSyntaxString(null,null));
+ System.out.println(test);
+
+// s.add(outer.toKBSyntaxString(null,null));
+ SparqlQueryDescriptionConvertVisitor testVisitor = new SparqlQueryDescriptionConvertVisitor();
+ String q = testVisitor.getSparqlQuery(outer.toKBSyntaxString());
+ System.out.println(q);
+ if (true) {
+ System.exit(0);
+ }
+
// <http://nlp2rdf.org/ontology/sentencefinalpunctuation_tag>
String query = "";
SparqlQueryDescriptionConvertVisitor visit = new SparqlQueryDescriptionConvertVisitor();
- visit.setLabels(true);
- visit.setDistinct(true);
- visit.setClassToSubclassesVirtuoso(subclassMap);
+ visit.setLabels(false);
+ visit.setDistinct(false);
+// visit.setClassToSubclassesVirtuoso(subclassMap);
+
+
for (String kbsyntax : s) {
query = visit.getSparqlQuery(kbsyntax);
result.put(kbsyntax, query);
Modified: trunk/src/dl-learner/org/dllearner/scripts/tiger/ExperimentConfig.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/tiger/ExperimentConfig.java 2010-02-13 17:45:31 UTC (rev 2033)
+++ trunk/src/dl-learner/org/dllearner/scripts/tiger/ExperimentConfig.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -8,11 +8,12 @@
public int resultLimit = -1;
public int splits = 5;
- public int initialsplits = 10;
- public int iteration = 5;
+ public int initialsplits = 30;
+ public int iteration = 1;
public boolean useStartClass = true;
- public int noise = 5;
+ public boolean searchTree = false;
+ public int noise = 0;
//sets ValueFrequency treshold and maxExecution time
public boolean adaptive = true;
public int maxExecutionTime = 40;
Added: trunk/src/dl-learner/org/dllearner/scripts/tiger/GlobalTest.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/tiger/GlobalTest.java (rev 0)
+++ trunk/src/dl-learner/org/dllearner/scripts/tiger/GlobalTest.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -0,0 +1,407 @@
+package org.dllearner.scripts.tiger;
+
+import java.io.BufferedReader;
+import java.io.File;
+import java.io.FileInputStream;
+import java.io.IOException;
+import java.io.InputStreamReader;
+import java.net.URL;
+import java.text.DecimalFormat;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Set;
+import java.util.SortedSet;
+import java.util.TreeSet;
+
+import org.apache.log4j.Level;
+import org.apache.log4j.Logger;
+import org.dllearner.algorithms.refinement2.ROLComponent2;
+import org.dllearner.algorithms.refinement2.ROLearner2;
+import org.dllearner.core.ComponentPool;
+import org.dllearner.core.EvaluatedDescription;
+import org.dllearner.core.KnowledgeSource;
+import org.dllearner.core.LearningAlgorithm;
+import org.dllearner.core.LearningProblem;
+import org.dllearner.core.ReasonerComponent;
+import org.dllearner.core.configurators.ComponentFactory;
+import org.dllearner.core.owl.Description;
+import org.dllearner.core.owl.Individual;
+import org.dllearner.kb.OWLFile;
+import org.dllearner.kb.sparql.Cache;
+import org.dllearner.kb.sparql.SPARQLTasks;
+import org.dllearner.kb.sparql.SparqlEndpoint;
+import org.dllearner.kb.sparql.SparqlKnowledgeSource;
+import org.dllearner.kb.sparql.SparqlQuery;
+import org.dllearner.kb.sparql.SparqlQueryDescriptionConvertVisitor;
+import org.dllearner.learningproblems.PosNegLPStandard;
+import org.dllearner.reasoning.FastInstanceChecker;
+import org.dllearner.refinementoperators.RhoDRDown;
+import org.dllearner.utilities.Helper;
+import org.dllearner.utilities.JamonMonitorLogger;
+import org.dllearner.utilities.examples.ExMakerFixedSize;
+import org.dllearner.utilities.examples.ExampleDataCollector;
+import org.dllearner.utilities.examples.Examples;
+
+import com.jamonapi.Monitor;
+
+public class GlobalTest {
+ private static final Logger logger = Logger.getLogger(GlobalTest.class);
+
+ static DecimalFormat df = new DecimalFormat("00.###%");
+
+ static String backgroundXML = "files/tiger.noSchema.noImports.rdf";
+ static String propertiesXML = "files/propertiesOnly.rdf";
+ static String sentenceXMLFolder = "files/tiger/";
+ static String sentenceprefix = "http://nlp2rdf.org/ontology/s";
+ static String prefix = "http://nlp2rdf.org/ontology/";
+
+ static String active = "files/active_all_sentenceNumbers.txt";
+ static String passiveNoZU = "files/passive_noZuInf_sentenceNumbers.txt";
+ static String passiveWithZu = "files/passive_zuInf_sentenceNumbers.txt";
+ static String test_has_pos = "files/test_has_pos.txt";
+ static String test_has_neg = "files/test_has_neg.txt";
+
+ static SparqlEndpoint sparqlEndpoint;
+ static SPARQLTasks sparqlTasks;
+
+ static String sparqlEndpointURL = "http://db0.aksw.org:8893/sparql";
+ static String graph = "http://nlp2rdf.org/tiger";
+ static String rulegraph = "http://nlp2rdf.org/schema/rules1";
+
+
+
+
+
+ final static boolean debug = false;
+ //no randomization in examples
+ final static boolean randomizedebug = !debug;
+
+ public static void main(String[] args) {
+ LogHelper.initLoggers();
+ Logger.getLogger(Cache.class).setLevel(Level.INFO);
+ Logger.getLogger(ComponentPool.class).setLevel(Level.INFO);
+ Logger.getLogger(ROLearner2.class).setLevel(Level.TRACE);
+ Logger.getLogger(RhoDRDown.class).setLevel(Level.TRACE);
+ Logger.getLogger(SparqlQuery.class).setLevel(Level.INFO);
+
+ try {
+ sparqlEndpoint = new SparqlEndpoint(new URL(sparqlEndpointURL), new ArrayList<String>(Arrays
+ .asList(new String[] { graph })), new ArrayList<String>());
+ sparqlTasks = new SPARQLTasks(Cache.getDefaultCache(), sparqlEndpoint);
+ } catch (Exception e) {
+ e.printStackTrace();
+ }
+
+ Examples allExamples = new Examples();
+ SortedSet<String> positives;
+ SortedSet<String> negatives;
+
+// positives = read(passiveWithZu);
+ positives = read(passiveNoZU);
+ negatives = read(active);
+
+ //removing overlap
+ positives.removeAll(negatives);
+ negatives.removeAll(positives);
+
+// System.out.println(Helper.intersection(passiveZuInfSentences, activeSentences));
+// System.out.println(Helper.intersection(passiveZuInfSentences, passiveNoZuSentences));
+// System.out.println(Helper.intersection(activeSentences, passiveNoZuSentences));
+ allExamples.addPosTrain(positives);
+ allExamples.addNegTrain(negatives);
+
+ logger.debug("All examples \n"+allExamples);
+
+ ExperimentConfig config = new ExperimentConfig();
+ firstContact( allExamples, config);
+ JamonMonitorLogger.writeHTMLReport("log/tiger.html");
+ //retrieved wird neues Example, als schnittmenge mit all
+ //und den bisher gewaehlten
+ //dann splits auswählen und
+ //pos und neg wieder hinzufuegen
+
+ }
+
+ public static void firstContact(Examples allExamples, ExperimentConfig config){
+ ExMakerFixedSize fs = new ExMakerFixedSize(allExamples, randomizedebug);
+ Examples learn = fs.select(config.initialsplits, config.initialsplits);
+ logger.debug("Intial training set \n"+learn);
+// System.out.println(learn.getPosTrain());
+// System.out.println(learn.getNegTrain());
+// if (true) {
+// System.exit(0);
+// }
+// int size = 0;
+ for(int i = 0 ; config.stopCondition(i, learn) ;i++ ) {
+ /*LEARNING*/
+ EvaluatedDescription ed = learn(learn, config);
+
+ /*RETRIEVING*/
+ SortedSet<String> retrieved = getSentences(ed, config.resultLimit);
+ logger.debug("Retrieved "+retrieved.size()+" sentences");
+
+
+ /*MASHING*/
+ //Menge aller positiven geschn. mit den gefundenen
+ SortedSet<String> posAsPos = Helper.intersection(retrieved, allExamples.getPosTrain());
+ logger.debug("Number of retrieved positives: "+posAsPos.size());
+ logger.debug("Number of total positives: "+allExamples.getPosTrain().size());
+ results(posAsPos, retrieved, allExamples);
+
+ //Menge aller positiven geschn. mit den gefundenen
+ SortedSet<String> negAsPos = Helper.intersection(retrieved, allExamples.getNegTrain());
+ logger.debug("Number of retrieved negatives: "+negAsPos.size());
+ logger.debug("Total: "+posAsPos.size()+" + "+negAsPos.size() +" = "+retrieved.size());
+
+// if(retrieved.size()!=(posAsPos.size()+negAsPos.size())){
+// logger.warn("sets are wrong");
+// System.exit(0);
+// }
+
+ Examples newlyFound = new Examples();
+ newlyFound.addPosTrain(Helper.intersection(retrieved, learn.getPosTest()));
+ newlyFound.addNegTrain(Helper.intersection(retrieved, learn.getNegTest()));
+ //validate here
+
+ fs = new ExMakerFixedSize(newlyFound, randomizedebug);
+ newlyFound = fs.select(config.splits, config.splits);
+
+ learn.addPosTrain(newlyFound.getPosTrain());
+ learn.addNegTrain(newlyFound.getNegTrain());
+ logger.debug("Next training set \n"+learn);
+// size = learn.getPosTrain().size() + learn.getNegTrain().size();
+
+ }
+
+
+
+
+
+ }
+
+ private static void results(SortedSet<String> posAsPos, SortedSet<String> retrieved, Examples allExamples) {
+ double precision = precision( posAsPos.size(), retrieved.size());
+ double recall = recall( posAsPos.size(),allExamples.getPosTrain().size());
+ logger.info("F-Measure: "+df.format( (2*precision*recall)/(precision+recall)) );
+
+ }
+
+ public static double precision( int posAsPos, int retrieved){
+ double precision = ((double)posAsPos)/((double)retrieved);
+ logger.info("Precision: "+df.format(precision));
+ return precision;
+ }
+ public static double recall( int posAsPos, int allPositives){
+ double recall = ((double)posAsPos)/((double)allPositives);
+
+ logger.info("Recall: "+df.format(recall));
+ return recall;
+
+ }
+
+ private static Set<KnowledgeSource> _getOWL(Examples ex) throws Exception{
+ Set<KnowledgeSource> tmp = new HashSet<KnowledgeSource>();
+ List<URL> urls = new ArrayList<URL>();
+ urls.addAll(ExampleDataCollector.convert(sentenceXMLFolder, ex.getPosTrain()));
+ urls.addAll(ExampleDataCollector.convert(sentenceXMLFolder, ex.getNegTrain()));
+ urls.add(new File(backgroundXML).toURI().toURL());
+
+ for (URL u : urls) {
+ OWLFile ks = ComponentFactory.getOWLFile(u);
+ tmp.add(ks);
+ }
+ return tmp;
+ }
+ @SuppressWarnings("unused")
+ private static Set<KnowledgeSource> _getSPARQL(Examples ex) throws Exception{
+ Set<KnowledgeSource> tmp = new HashSet<KnowledgeSource>();
+
+ Set<String> examples = new TreeSet<String>();
+ examples.addAll(ex.getPosTrain());
+ examples.addAll(ex.getNegTrain());
+ SparqlKnowledgeSource ks = ComponentFactory.getSparqlKnowledgeSource(new URL(sparqlEndpointURL), examples);
+ ks.getConfigurator().setUrl(new URL(sparqlEndpointURL));
+ ks.getConfigurator().setDefaultGraphURIs(new HashSet<String>(Arrays.asList(new String[]{graph})));
+ ks.getConfigurator().setInstances(examples);
+ ks.getConfigurator().setDissolveBlankNodes(false);
+ ks.getConfigurator().setRecursionDepth(2);
+ ks.getConfigurator().setDissolveBlankNodes(false);
+ ks.getConfigurator().setCloseAfterRecursion(true);
+ ks.getConfigurator().setGetAllSuperClasses(true);
+ ks.getConfigurator().setGetPropertyInformation(false);
+ ks.getConfigurator().setUseLits(true);
+// ks.getConfigurator().
+ OWLFile ks2 = ComponentFactory.getOWLFile(new File(propertiesXML).toURI().toURL());
+ tmp.add(ks);
+ tmp.add(ks2);
+
+ return tmp;
+ }
+
+ //test if virtuoso is correct
+ public static void validate(Description d, Examples newlyFound){
+ try {
+ ExMakerFixedSize fs = new ExMakerFixedSize(newlyFound);
+ Examples tmp = fs.select(100, 100);
+ FastInstanceChecker fc = _getFastInstanceChecker(tmp);
+ SortedSet<Individual> inds = fc.getIndividuals(d);
+ } catch (Exception e) {
+ e.printStackTrace();
+ }
+ }
+
+ public static FastInstanceChecker _getFastInstanceChecker(Examples ex)throws Exception{
+ Set<KnowledgeSource> tmp = _getOWL(ex);
+// Set<KnowledgeSource> tmp = _getSPARQL(ex);
+
+
+ FastInstanceChecker rc = ComponentFactory.getFastInstanceChecker(tmp);
+ for (KnowledgeSource ks : tmp) {
+ ks.init();
+ }
+ rc.init();
+ return rc;
+ }
+
+ public static EvaluatedDescription learn(Examples ex, ExperimentConfig config) {
+ Monitor init = JamonMonitorLogger.getTimeMonitor(GlobalTest.class, "init").start();
+
+ EvaluatedDescription result = null;
+
+ try {
+ FastInstanceChecker rc = _getFastInstanceChecker(ex);
+ PosNegLPStandard lp = ComponentFactory
+ .getPosNegLPStandard(rc, ex.getPosTrain(), ex.getNegTrain());
+ LearningAlgorithm la = _getROLLearner(lp, rc, config, ex);
+ lp.init();
+ la.init();
+ init.stop();
+ Monitor learning = JamonMonitorLogger.getTimeMonitor(GlobalTest.class, "learning")
+ .start();
+ la.start();
+ learning.stop();
+
+ result = la.getCurrentlyBestEvaluatedDescription();
+ logger.debug(PrefixMap.toKBSyntaxString(result.getDescription()));
+ logger.debug(PrefixMap.toManchesterSyntaxString(result.getDescription()));
+
+ } catch (Exception e) {
+ e.printStackTrace();
+ System.exit(0);
+ }
+ return result;
+ }
+
+ public static SortedSet<String> getSentences(EvaluatedDescription ed, int resultLimit) {
+ SortedSet<String> result = new TreeSet<String>();
+ SparqlQueryDescriptionConvertVisitor visit = new SparqlQueryDescriptionConvertVisitor();
+ visit.setDistinct(true);
+ visit.setLabels(false);
+ visit.setLimit(resultLimit);
+ String sparqlQuery = "";
+ try {
+ sparqlQuery = visit.getSparqlQuery(ed.getDescription());
+ } catch (Exception e1) {
+ e1.printStackTrace();
+ }
+ logger.debug(PrefixMap.toKBSyntaxString(ed.getDescription()));
+ sparqlQuery = " \n define input:inference \"" + rulegraph + "\" \n" + "" + sparqlQuery;
+ logger.debug(sparqlQuery);
+
+ Monitor m = JamonMonitorLogger.getTimeMonitor(GlobalTest.class, "sparqlquery").start();
+ result.addAll(sparqlTasks.queryAsSet(sparqlQuery, "subject"));
+ m.stop();
+ logger.debug("query avg: " + ((double)m.getAvg() / (double)1000)+ " seconds (last: "+((double)m.getLastValue() / (double)1000)+")");
+ if(result.isEmpty()){
+
+ logger.error("sparql query returned no results ");
+ logger.error(sparqlQuery);
+ System.exit(0);
+ }
+ return result;
+ }
+
+ private static LearningAlgorithm _getROLLearner(LearningProblem lp, ReasonerComponent rc, ExperimentConfig config, Examples ex)
+ throws Exception {
+
+ int maxExecutionTime = config.maxExecutionTime;
+ int valueFrequencyThreshold = config.valueFrequencyThreshold;
+ if(config.adaptive){
+ maxExecutionTime = 2 * ex.sizeOfTrainingSets();
+ valueFrequencyThreshold = ex.getPosTrain().size();
+// valueFrequencyThreshold = (int) Math.floor(0.8d*((double)ex.getPosTrain().size()));
+
+ }
+
+ ROLComponent2 la = ComponentFactory.getROLComponent2(lp, rc);
+ la.getConfigurator().setUseExistsConstructor(true);
+
+ la.getConfigurator().setUseAllConstructor(false);
+ la.getConfigurator().setUseCardinalityRestrictions(false);
+ la.getConfigurator().setUseNegation(false);
+ la.getConfigurator().setUseHasValueConstructor(false);
+ la.getConfigurator().setUseDataHasValueConstructor(true);
+ la.getConfigurator().setValueFrequencyThreshold(valueFrequencyThreshold);
+
+ la.getConfigurator().setIgnoredConcepts(new HashSet<String>(Arrays.asList(new String[]{
+ "http://nlp2rdf.org/ontology/sentencefinalpunctuation_tag",
+ "http://nlp2rdf.org/ontology/comma_tag",
+ "http://nachhalt.sfb632.uni-potsdam.de/owl/stts.owl#SentenceFinalPunctuation"
+ })));
+
+
+ la.getConfigurator().setNoisePercentage(config.noise);
+ la.getConfigurator().setTerminateOnNoiseReached(true);
+ la.getConfigurator().setMaxExecutionTimeInSeconds(maxExecutionTime);
+
+ if(config.useStartClass){
+ la.getConfigurator().setStartClass(prefix + "Sentence");
+ }
+
+ la.getConfigurator().setWriteSearchTree(config.searchTree);
+ la.getConfigurator().setSearchTreeFile("log/searchTreeTiger.txt");
+ la.getConfigurator().setReplaceSearchTree(true);
+ return la;
+ }
+
+ public static SortedSet<String> read(String f) {
+ SortedSet<String> result = new TreeSet<String>();
+ BufferedReader in = null;
+ try {
+ in = new BufferedReader(new InputStreamReader(new FileInputStream(f)));
+
+ String line;
+ while ((line = in.readLine()) != null) {
+ try {
+ line = line.trim();
+ Integer.parseInt(line);
+ if (!result.add(sentenceprefix + line)) {
+ logger.error("reading failed");
+ System.exit(0);
+ }
+ } catch (Exception e) {
+ e.printStackTrace();
+ System.exit(0);
+ }
+ }
+
+ } catch (Exception e) {
+ e.printStackTrace();
+ logger.error("Could not read examples from: " + f);
+ System.exit(0);
+
+ } finally {
+ try {
+ in.close();
+ } catch (IOException e) {
+ e.printStackTrace();
+ }
+ }
+ logger.info("read " + result.size() + " lines from " + f);
+
+ return result;
+ }
+
+}
Modified: trunk/src/dl-learner/org/dllearner/scripts/tiger/TestIterativeLearning.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/tiger/TestIterativeLearning.java 2010-02-13 17:45:31 UTC (rev 2033)
+++ trunk/src/dl-learner/org/dllearner/scripts/tiger/TestIterativeLearning.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -26,6 +26,8 @@
import org.dllearner.core.LearningProblem;
import org.dllearner.core.ReasonerComponent;
import org.dllearner.core.configurators.ComponentFactory;
+import org.dllearner.core.owl.Description;
+import org.dllearner.core.owl.Individual;
import org.dllearner.kb.OWLFile;
import org.dllearner.kb.sparql.Cache;
import org.dllearner.kb.sparql.SPARQLTasks;
@@ -36,6 +38,7 @@
import org.dllearner.learningproblems.PosNegLPStandard;
import org.dllearner.reasoning.FastInstanceChecker;
import org.dllearner.refinementoperators.RhoDRDown;
+import org.dllearner.utilities.Files;
import org.dllearner.utilities.Helper;
import org.dllearner.utilities.JamonMonitorLogger;
import org.dllearner.utilities.examples.ExMakerFixedSize;
@@ -47,7 +50,10 @@
public class TestIterativeLearning {
private static final Logger logger = Logger.getLogger(TestIterativeLearning.class);
- static String backgroundXML = "files/tiger.noSchema.noImports.rdf";
+ static DecimalFormat df = new DecimalFormat("00.###%");
+
+// static String backgroundXML = "files/tiger.noSchema.noImports.rdf";
+ static String backgroundXML = "files/tiger_trimmed_toPOS.rdf";
static String propertiesXML = "files/propertiesOnly.rdf";
static String sentenceXMLFolder = "files/tiger/";
static String sentenceprefix = "http://nlp2rdf.org/ontology/s";
@@ -68,6 +74,8 @@
+
+
final static boolean debug = false;
//no randomization in examples
final static boolean randomizedebug = !debug;
@@ -133,7 +141,7 @@
EvaluatedDescription ed = learn(learn, config);
/*RETRIEVING*/
- SortedSet<String> retrieved = getSentences(ed, config.resultLimit);
+ SortedSet<String> retrieved = getSentences(ed, config.resultLimit, learn);
logger.debug("Retrieved "+retrieved.size()+" sentences");
@@ -142,23 +150,34 @@
SortedSet<String> posAsPos = Helper.intersection(retrieved, allExamples.getPosTrain());
logger.debug("Number of retrieved positives: "+posAsPos.size());
logger.debug("Number of total positives: "+allExamples.getPosTrain().size());
- precision( posAsPos.size(), retrieved.size());
+ results(posAsPos, retrieved, allExamples);
//Menge aller positiven geschn. mit den gefundenen
- recall( posAsPos.size(),allExamples.getPosTrain().size());
-
SortedSet<String> negAsPos = Helper.intersection(retrieved, allExamples.getNegTrain());
logger.debug("Number of retrieved negatives: "+negAsPos.size());
logger.debug("Total: "+posAsPos.size()+" + "+negAsPos.size() +" = "+retrieved.size());
+ //not covered
+
+
// if(retrieved.size()!=(posAsPos.size()+negAsPos.size())){
// logger.warn("sets are wrong");
// System.exit(0);
// }
Examples newlyFound = new Examples();
- newlyFound.addPosTrain(Helper.intersection(retrieved, learn.getPosTest()));
- newlyFound.addNegTrain(Helper.intersection(retrieved, learn.getNegTest()));
+ SortedSet<String> discoveredPosInStore = Helper.intersection(retrieved, learn.getPosTest());
+ SortedSet<String> misclassifiedNegInStore = Helper.intersection(retrieved, learn.getNegTest());
+ newlyFound.addPosTrain(discoveredPosInStore);
+ newlyFound.addNegTrain(misclassifiedNegInStore);
+ int print = 5;
+ logger.info("Discovered "+discoveredPosInStore.size()+" positive sentences in store (printing "+print+"):");
+ _getLabels(discoveredPosInStore, print);
+ logger.info("Misclassified "+misclassifiedNegInStore.size()+" negative sentences in store (printing "+print+"):");
+ _getLabels(misclassifiedNegInStore, print);
+
+
+
fs = new ExMakerFixedSize(newlyFound, randomizedebug);
newlyFound = fs.select(config.splits, config.splits);
@@ -175,20 +194,33 @@
}
- public static void precision( int posAsPos, int retrieved){
- logger.info("Precision: "+DecimalFormat.getPercentInstance().format(((double)posAsPos)/(double)retrieved)+"%");
+ private static void results(SortedSet<String> posAsPos, SortedSet<String> retrieved, Examples allExamples) {
+ double precision = precision( posAsPos.size(), retrieved.size());
+ double recall = recall( posAsPos.size(),allExamples.getPosTrain().size());
+ logger.info("F-Measure: "+df.format( (2*precision*recall)/(precision+recall)) );
+
}
- public static void recall( int posAsPos, int allPositives){
- logger.info("Recall: "+DecimalFormat.getPercentInstance().format(((double)posAsPos)/(double)allPositives)+"%");
+
+ public static double precision( int posAsPos, int retrieved){
+ double precision = ((double)posAsPos)/((double)retrieved);
+ logger.info("Precision: "+df.format(precision));
+ return precision;
+ }
+ public static double recall( int posAsPos, int allPositives){
+ double recall = ((double)posAsPos)/((double)allPositives);
+ logger.info("Recall: "+df.format(recall));
+ return recall;
+
}
private static Set<KnowledgeSource> _getOWL(Examples ex) throws Exception{
Set<KnowledgeSource> tmp = new HashSet<KnowledgeSource>();
List<URL> urls = new ArrayList<URL>();
+ urls.add(new File(backgroundXML).toURI().toURL());
urls.addAll(ExampleDataCollector.convert(sentenceXMLFolder, ex.getPosTrain()));
urls.addAll(ExampleDataCollector.convert(sentenceXMLFolder, ex.getNegTrain()));
- urls.add(new File(backgroundXML).toURI().toURL());
+
for (URL u : urls) {
OWLFile ks = ComponentFactory.getOWLFile(u);
@@ -222,25 +254,42 @@
return tmp;
}
- public static EvaluatedDescription learn(Examples ex, ExperimentConfig config) {
- Monitor init = JamonMonitorLogger.getTimeMonitor(TestIterativeLearning.class, "init").start();
-
- EvaluatedDescription result = null;
-
+ //test if virtuoso is correct
+ public static void validate(Description d, Examples newlyFound){
try {
+ ExMakerFixedSize fs = new ExMakerFixedSize(newlyFound);
+ Examples tmp = fs.select(100, 100);
+ FastInstanceChecker fc = _getFastInstanceChecker(tmp);
+ @SuppressWarnings("unused")
+ SortedSet<Individual> inds = fc.getIndividuals(d);
+ } catch (Exception e) {
+ e.printStackTrace();
+ }
+ }
+
+ public static FastInstanceChecker _getFastInstanceChecker(Examples ex)throws Exception{
Set<KnowledgeSource> tmp = _getOWL(ex);
// Set<KnowledgeSource> tmp = _getSPARQL(ex);
FastInstanceChecker rc = ComponentFactory.getFastInstanceChecker(tmp);
- PosNegLPStandard lp = ComponentFactory
- .getPosNegLPStandard(rc, ex.getPosTrain(), ex.getNegTrain());
- LearningAlgorithm la = _getROLLearner(lp, rc, config, ex);
-
for (KnowledgeSource ks : tmp) {
ks.init();
}
rc.init();
+ return rc;
+ }
+
+ public static EvaluatedDescription learn(Examples ex, ExperimentConfig config) {
+ Monitor init = JamonMonitorLogger.getTimeMonitor(TestIterativeLearning.class, "init").start();
+
+ EvaluatedDescription result = null;
+
+ try {
+ FastInstanceChecker rc = _getFastInstanceChecker(ex);
+ PosNegLPStandard lp = ComponentFactory
+ .getPosNegLPStandard(rc, ex.getPosTrain(), ex.getNegTrain());
+ LearningAlgorithm la = _getROLLearner(lp, rc, config, ex);
lp.init();
la.init();
init.stop();
@@ -260,33 +309,64 @@
return result;
}
- public static SortedSet<String> getSentences(EvaluatedDescription ed, int resultLimit) {
+ public static SortedSet<String> getSentences(EvaluatedDescription ed, int resultLimit, Examples justforFindingTheBug) {
SortedSet<String> result = new TreeSet<String>();
SparqlQueryDescriptionConvertVisitor visit = new SparqlQueryDescriptionConvertVisitor();
visit.setDistinct(true);
visit.setLabels(false);
visit.setLimit(resultLimit);
- String sparqlQuery = "";
+ String sparqlQueryGood = "";
+ String sparqlQueryBad = "";
try {
- sparqlQuery = visit.getSparqlQuery(ed.getDescription());
+ sparqlQueryGood = visit.getSparqlQuery(ed.getDescription().toKBSyntaxString());
+ sparqlQueryBad = visit.getSparqlQuery(ed.getDescription());
+ if(!sparqlQueryGood.equals(sparqlQueryBad)){
+ String file = "errorDescription/"+System.currentTimeMillis();
+ justforFindingTheBug.writeExamples(file);
+ Files.appendFile(new File(file), "\n\n/**\nGood:\n"+sparqlQueryGood+"\nBad:\n"+sparqlQueryBad+"**/");
+ }
+
} catch (Exception e1) {
e1.printStackTrace();
}
- logger.debug(PrefixMap.toKBSyntaxString(ed.getDescription()));
- sparqlQuery = " \n define input:inference \"" + rulegraph + "\" \n" + "" + sparqlQuery;
- logger.debug(sparqlQuery);
+ logger.debug("USING CONCEPT: "+PrefixMap.toKBSyntaxString(ed.getDescription()));
+ sparqlQueryGood = " \n define input:inference \"" + rulegraph + "\" \n" + "" + sparqlQueryGood;
+ logger.debug(sparqlQueryGood);
Monitor m = JamonMonitorLogger.getTimeMonitor(TestIterativeLearning.class, "sparqlquery").start();
- result.addAll(sparqlTasks.queryAsSet(sparqlQuery, "subject"));
+ result.addAll(sparqlTasks.queryAsSet(sparqlQueryGood, "subject"));
+ m.stop();
logger.debug("query avg: " + ((double)m.getAvg() / (double)1000)+ " seconds (last: "+((double)m.getLastValue() / (double)1000)+")");
if(result.isEmpty()){
logger.error("sparql query returned no results ");
- logger.error(sparqlQuery);
+ logger.error(sparqlQueryGood);
System.exit(0);
}
return result;
}
+
+ private static void _getLabels(SortedSet<String> sentenceURIs, int limit){
+ int i = 0;
+ for (String sentenceURI : sentenceURIs) {
+ if(i>=limit){
+ break;
+ }
+ i++;
+ _getLabel(sentenceURI);
+ }
+ }
+
+ private static void _getLabel(String sentenceURI){
+ String query = "SELECT * FROM <"+graph+"> " +
+ "{ <"+sentenceURI+"> rdfs:label ?label . }";
+ SortedSet<String> s = sparqlTasks.queryAsSet(query, "label");
+ if(s.isEmpty()){
+ logger.warn("no label for "+sentenceURI);
+ }else{
+ logger.debug(sentenceURI.replace(prefix, "")+" "+s.first());
+ }
+ }
private static LearningAlgorithm _getROLLearner(LearningProblem lp, ReasonerComponent rc, ExperimentConfig config, Examples ex)
throws Exception {
@@ -295,7 +375,8 @@
int valueFrequencyThreshold = config.valueFrequencyThreshold;
if(config.adaptive){
maxExecutionTime = 2 * ex.sizeOfTrainingSets();
- valueFrequencyThreshold = (int) Math.floor(0.8d*((double)ex.getPosTrain().size()));
+ valueFrequencyThreshold = ex.getPosTrain().size();
+// valueFrequencyThreshold = (int) Math.floor(0.8d*((double)ex.getPosTrain().size()));
}
@@ -324,9 +405,9 @@
la.getConfigurator().setStartClass(prefix + "Sentence");
}
- la.getConfigurator().setWriteSearchTree(false);
- la.getConfigurator().setSearchTreeFile("log/searchTree.txt");
- la.getConfigurator().setReplaceSearchTree(false);
+ la.getConfigurator().setWriteSearchTree(config.searchTree);
+ la.getConfigurator().setSearchTreeFile("log/searchTreeTiger.txt");
+ la.getConfigurator().setReplaceSearchTree(true);
return la;
}
Modified: trunk/src/dl-learner/org/dllearner/utilities/examples/Examples.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/utilities/examples/Examples.java 2010-02-13 17:45:31 UTC (rev 2033)
+++ trunk/src/dl-learner/org/dllearner/utilities/examples/Examples.java 2010-02-13 19:33:50 UTC (rev 2034)
@@ -26,6 +26,7 @@
import java.util.TreeSet;
import org.apache.log4j.Logger;
+import org.dllearner.utilities.URLencodeUTF8;
/**
* a container for examples
@@ -154,12 +155,19 @@
public void writeExamples(String filename) {
try {
FileWriter a = new FileWriter(filename, false);
+
+ StringBuffer buffer = new StringBuffer();
+ buffer.append("\n\n\n\n\n");
for (String s : posTrain) {
- a.write("+\"" + s + "\"\n");
+ a.write("import(\""+URLencodeUTF8.encode(s)+"\");\n");
+ buffer.append("+\"" + s + "\"\n");
}
for (String s : negTrain) {
- a.write("-\"" + s + "\"\n");
+ a.write("import(\""+URLencodeUTF8.encode(s)+"\");\n");
+ buffer.append("-\"" + s + "\"\n");
}
+
+ a.write(buffer.toString());
a.flush();
a.close();
logger.info("wrote examples to " + filename);
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
