Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner:[2665] trunk
|
From: <lor...@us...> - 2011-02-26 11:49:36
|
Revision: 2665
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=2665&view=rev
Author: lorenz_b
Date: 2011-02-26 11:49:29 +0000 (Sat, 26 Feb 2011)
Log Message:
-----------
Continued evaluation script.
Added class to preprocess given question by extracting words, remove stop words and stem them.
Modified Paths:
--------------
trunk/autosparql/src/main/java/org/dllearner/autosparql/evaluation/EvaluationWithNLQueriesScript.java
trunk/autosparql/src/main/java/org/dllearner/autosparql/server/ExampleFinder.java
trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/DBpediaLuceneIndexCreator.java
trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/LuceneSearch.java
trunk/sparql-query-generator/src/main/java/org/dllearner/sparqlquerygenerator/util/QuestionBasedStatementFilter.java
Added Paths:
-----------
trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/QuestionProcessor.java
Modified: trunk/autosparql/src/main/java/org/dllearner/autosparql/evaluation/EvaluationWithNLQueriesScript.java
===================================================================
--- trunk/autosparql/src/main/java/org/dllearner/autosparql/evaluation/EvaluationWithNLQueriesScript.java 2011-02-25 11:23:58 UTC (rev 2664)
+++ trunk/autosparql/src/main/java/org/dllearner/autosparql/evaluation/EvaluationWithNLQueriesScript.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -5,12 +5,15 @@
import java.io.BufferedReader;
import java.io.File;
+import java.io.FileReader;
import java.io.IOException;
import java.io.InputStreamReader;
+import java.io.StringReader;
import java.net.MalformedURLException;
import java.net.URL;
import java.net.URLConnection;
import java.util.ArrayList;
+import java.util.Arrays;
import java.util.Collections;
import java.util.HashSet;
import java.util.Hashtable;
@@ -25,9 +28,17 @@
import javax.xml.parsers.ParserConfigurationException;
import org.apache.log4j.ConsoleAppender;
+import org.apache.log4j.FileAppender;
+import org.apache.log4j.Layout;
import org.apache.log4j.Level;
import org.apache.log4j.Logger;
import org.apache.log4j.PatternLayout;
+import org.apache.lucene.analysis.StopAnalyzer;
+import org.apache.lucene.analysis.StopFilter;
+import org.apache.lucene.analysis.Tokenizer;
+import org.apache.lucene.analysis.standard.StandardTokenizer;
+import org.apache.lucene.util.Version;
+import org.apache.solr.analysis.StopFilterFactory;
import org.apache.solr.client.solrj.SolrServer;
import org.apache.solr.client.solrj.SolrServerException;
import org.apache.solr.client.solrj.impl.CommonsHttpSolrServer;
@@ -37,13 +48,18 @@
import org.dllearner.autosparql.client.exception.SPARQLQueryException;
import org.dllearner.autosparql.server.ExampleFinder;
import org.dllearner.autosparql.server.Generalisation;
+import org.dllearner.autosparql.server.NBR;
import org.dllearner.autosparql.server.exception.TimeOutException;
+import org.dllearner.autosparql.server.search.DBpediaSchemaIndex;
+import org.dllearner.autosparql.server.search.LuceneSearch;
+import org.dllearner.autosparql.server.search.QuestionProcessor;
import org.dllearner.autosparql.server.util.SPARQLEndpointEx;
import org.dllearner.kb.sparql.ExtractionDBCache;
import org.dllearner.kb.sparql.SparqlEndpoint;
import org.dllearner.kb.sparql.SparqlQuery;
import org.dllearner.sparqlquerygenerator.operations.lgg.LGGGeneratorImpl;
import org.dllearner.sparqlquerygenerator.util.ExactMatchFilter;
+import org.dllearner.sparqlquerygenerator.util.QuestionBasedStatementFilter;
import org.w3c.dom.DOMException;
import org.w3c.dom.Document;
import org.w3c.dom.Element;
@@ -52,12 +68,19 @@
import com.hp.hpl.jena.query.ResultSet;
+import edu.stanford.nlp.ling.HasWord;
+import edu.stanford.nlp.ling.Sentence;
+import edu.stanford.nlp.ling.TaggedWord;
+import edu.stanford.nlp.tagger.maxent.MaxentTagger;
+
public class EvaluationWithNLQueriesScript {
private static Logger logger = Logger.getLogger(EvaluationWithNLQueriesScript.class);
private static final boolean USE_SYNONYMS = false;
private static final String SOLR_SERVER_URL = "http://139.18.2.164:8983/solr/dbpediaCore/";
private static final String QUERY_ANSWERS_FILE_PATH = "evaluation/dbpedia-train_cleaned.xml";
+ private static final String SCHEMA_FILE_PATH = "evaluation/dbpedia_schema.owl";
+ private static final String LUCENE_INDEX_DIRECTORY = "/opt/autosparql/index";
private static final SparqlEndpoint ENDPOINT = SparqlEndpoint.getEndpointDBpediaLiveAKSW();
private static final int NR_OF_POS_START_EXAMPLES_COUNT = 3;
@@ -76,7 +99,12 @@
private ExampleFinder exFinder;
+ private DBpediaSchemaIndex schemaIndex;
+ private LuceneSearch luceneSearch;
+ private QuestionProcessor qProcessor = new QuestionProcessor();
+
+
public EvaluationWithNLQueriesScript(){
try {
server = new CommonsHttpSolrServer(SOLR_SERVER_URL);
@@ -88,6 +116,8 @@
exFinder = new ExampleFinder(new SPARQLEndpointEx(
new SparqlEndpoint(new URL("http://lod.openlinksw.com/sparql"),
Collections.singletonList("http://dbpedia.org"), Collections.<String>emptyList()), null, null, predicateFilters), selectCache, constructCache);
+ schemaIndex = new DBpediaSchemaIndex(SCHEMA_FILE_PATH);
+ luceneSearch = new LuceneSearch(LUCENE_INDEX_DIRECTORY);
} catch (MalformedURLException e) {
e.printStackTrace();
}
@@ -137,13 +167,13 @@
logger.info("Done.");
}
- private Set<String> getResourcesByNLQuery(String query){
+ private Set<String> getResourcesByNLQuery(String question){
logger.info("Getting Top " + TOP_K + " resources related to question with Solr...");
Set<String> resources = new HashSet<String>();
QueryResponse response;
try {
ModifiableSolrParams params = new ModifiableSolrParams();
- params.set("q", query);
+ params.set("q", question);
params.set("rows", TOP_K);
response = server.query(params);
for(SolrDocument d : response.getResults()){
@@ -157,6 +187,14 @@
return resources;
}
+ private List<String> getRelevantWords(String question){
+ return qProcessor.getRelevantWords(question);
+// Properties props = new Properties();
+// props.put("annotators", "tokenize, ssplit, pos, lemma, ner, parse, dcoref");
+// StanfordCoreNLP pipeline = new StanfordCoreNLP(props);
+
+ }
+
private List<String> getResourcesByWikipedia(String query) {
logger.info("Getting Top " + TOP_K + " resources related to question with Wikipedia...");
long startTime = System.currentTimeMillis();
@@ -179,7 +217,6 @@
rd.close();
XMLPagesParser parser = new XMLPagesParser(sb.toString());
parser.parse();
- String title;
for (Page page : parser.getPagesList()) {
resources.add("http://dbpedia.org/resource/" + page.getTitle().replace(" ", "_"));
}
@@ -195,6 +232,15 @@
return resources;
}
+ private List<String> getSchemaElementsByQuery(String query){
+ logger.info("Getting Top " + TOP_K + " schema elements related to question with Lucene...");
+ long startTime = System.currentTimeMillis();
+ List<String> elements = schemaIndex.getResources(query);
+ logger.info("Got " + elements.size() + " elements in " + (System.currentTimeMillis()-startTime) + "ms.");
+ logger.info(elements);
+ return elements;
+ }
+
private Set<String> getResourcesBySPARQLQuery(String query){
logger.info("Sending query...");
long startTime = System.currentTimeMillis();
@@ -212,6 +258,7 @@
Set<String> answers;
List<String> examples;
Set<String> relatedResources;
+ List<String> relevantWords;
int i = 1;
for(String question : question2Answers.keySet()){
logger.info(getNewQuestionString(i++, question));
@@ -221,12 +268,23 @@
logger.info("Target query: \n" + targetQuery);
answers = question2Answers.get(question);
logger.info("Answers (" + answers.size() + "): " + answers);
- examples = getResourcesByWikipedia(question);
- relatedResources = getResourcesByNLQuery(question.substring(0, question.length()-1));
+ //preprocess question to extract only relevant words and set them as filter for statements
+ relevantWords = getRelevantWords(question);
+ exFinder.setStatementFilter(new QuestionBasedStatementFilter(new HashSet<String>(relevantWords)));
+ question = "";
+ for(String word : relevantWords){
+ question += " " + word;
+ }
+ question.trim();
+ logger.info("Rebuilt question string: " + question);
+ //get examples
+ examples = getResourcesByWikipedia(question);//luceneSearch.getResources(question)
+ //get resources which are relevant for query and add them as filter for objects
+// relatedResources = getResourcesByNLQuery(question.substring(0, question.length()-1));
+// relatedResources.addAll(getSchemaElementsByQuery(question.substring(0, question.length()-1)));
+// exFinder.setObjectFilter(new ExactMatchFilter(relatedResources));
- exFinder.setObjectFilter(new ExactMatchFilter(relatedResources));
-
- //select some positive example and negative examples
+ //select some positive and negative examples
List<String> posExamples = new ArrayList<String>();
List<String> negExamples = new ArrayList<String>();
for(String ex : examples){
@@ -303,8 +361,16 @@
Logger.getLogger(Generalisation.class).setLevel(Level.OFF);
Logger.getLogger(LGGGeneratorImpl.class).setLevel(Level.OFF);
Logger.getRootLogger().removeAllAppenders();
- ConsoleAppender appender = new ConsoleAppender(new PatternLayout("%m%n"));
+ Layout layout = new PatternLayout("%m%n");
+ ConsoleAppender appender = new ConsoleAppender(layout);
Logger.getRootLogger().addAppender(appender);
+ FileAppender fileAppender = new FileAppender(
+ layout, "log/evaluation.log", false);
+ fileAppender.setThreshold(Level.DEBUG);
+ Logger.getRootLogger().addAppender(fileAppender);
+ Logger.getLogger(NBR.class).setLevel(Level.DEBUG);
+
+
new EvaluationWithNLQueriesScript().evaluate();
}
Modified: trunk/autosparql/src/main/java/org/dllearner/autosparql/server/ExampleFinder.java
===================================================================
--- trunk/autosparql/src/main/java/org/dllearner/autosparql/server/ExampleFinder.java 2011-02-25 11:23:58 UTC (rev 2664)
+++ trunk/autosparql/src/main/java/org/dllearner/autosparql/server/ExampleFinder.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -30,6 +30,7 @@
import com.hp.hpl.jena.query.QuerySolution;
import com.hp.hpl.jena.query.ResultSetRewindable;
import com.hp.hpl.jena.rdf.model.Model;
+import com.hp.hpl.jena.rdf.model.Selector;
import com.hp.hpl.jena.vocabulary.RDFS;
public class ExampleFinder {
@@ -454,6 +455,10 @@
queryTreeCache.setObjectFilter(filter);
}
+ public void setStatementFilter(Selector filter){
+ queryTreeCache.setStatementFilter(filter);
+ }
+
public String getCurrentQuery(){
return currentQuery;
}
Modified: trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/DBpediaLuceneIndexCreator.java
===================================================================
--- trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/DBpediaLuceneIndexCreator.java 2011-02-25 11:23:58 UTC (rev 2664)
+++ trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/DBpediaLuceneIndexCreator.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -178,7 +178,7 @@
ps.setString(1, uri);
ResultSet rs = ps.executeQuery();
if(rs.next()){
- pageRank = rs.getInt("pagerank");
+ pageRank = rs.getInt("rank");
}
} catch (SQLException e) {
e.printStackTrace();
Modified: trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/LuceneSearch.java
===================================================================
--- trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/LuceneSearch.java 2011-02-25 11:23:58 UTC (rev 2664)
+++ trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/LuceneSearch.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -85,7 +85,7 @@
Document d = searcher.doc(doc.doc);
String uri = d.get("uri");
String label = d.get("label");
- String comment = d.get("abstract");
+ String comment = d.get("comment");
String imageURL = d.get("imageURL");
examples.add(new Example(uri, label, imageURL, comment));
}
Added: trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/QuestionProcessor.java
===================================================================
--- trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/QuestionProcessor.java (rev 0)
+++ trunk/autosparql/src/main/java/org/dllearner/autosparql/server/search/QuestionProcessor.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -0,0 +1,110 @@
+package org.dllearner.autosparql.server.search;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.StringReader;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+
+import org.apache.log4j.Logger;
+
+import edu.stanford.nlp.ling.HasWord;
+import edu.stanford.nlp.ling.TaggedWord;
+import edu.stanford.nlp.process.Morphology;
+import edu.stanford.nlp.tagger.maxent.MaxentTagger;
+
+public class QuestionProcessor {
+
+ private final Logger logger = Logger.getLogger(QuestionProcessor.class);
+
+ private MaxentTagger tagger;
+ private final List<String> stopWords = Arrays.asList(
+ "a", "all", "an", "and", "are", "as", "at", "be", "but", "by",
+ "for", "he", "if", "in", "into", "is", "it", "me",
+ "no", "not", "of", "on", "or", "she", "such",
+ "that", "the", "their", "then", "there", "these",
+ "they", "this", "to", "was", "will", "with"
+ );
+
+ public QuestionProcessor(){
+ try {
+ tagger = new MaxentTagger("src/main/resources/de/simba/ner/models/left3words-wsj-0-18.tagger");
+ } catch (IOException e) {
+ e.printStackTrace();
+ } catch (ClassNotFoundException e) {
+ e.printStackTrace();
+ }
+ }
+
+ public List<String> getRelevantWords(String question){
+ logger.info("Processing question \"" + question + "\"...");
+ //tokenize question
+ List<String> words = getWords(question);
+ logger.info("Extracted words: " + words);
+ //remove stop words
+ removeStopWords(words);
+ logger.info("After removed stop words: " + words);
+ //stem words
+ words = getStemmedWords(words);
+ logger.info("After stemming: " + words);
+
+ return words;
+ }
+
+ private List<String> getWords(String question){
+ List<String> words = new ArrayList<String>();
+ List<ArrayList<? extends HasWord>> sentences = tagger.tokenizeText(new BufferedReader(new StringReader(question)));
+ for (ArrayList<? extends HasWord> sentence : sentences) {
+ ArrayList<TaggedWord> tSentence = tagger.tagSentence(sentence);
+ String nounPhrase = "";
+ boolean firstWord = true;
+ for(TaggedWord tWord : tSentence){
+ //ignore first word if it is a verb
+ if(firstWord){
+ if(tWord.tag().startsWith("V")){
+ continue;
+ }
+ firstWord = false;
+ }
+ //if words belongs to noun phrase treat them as one single term
+ if(tWord.tag().equals("NNP")){
+ nounPhrase += " " + tWord.word();
+ } else {
+ if(!nounPhrase.isEmpty()){
+ words.add(nounPhrase.trim());
+ nounPhrase = "";
+ }
+ //ignore punctuation signs
+ if(!tWord.tag().equals(".")){
+ words.add(tWord.word());
+ }
+ }
+
+ }
+ if(!nounPhrase.isEmpty()){
+ words.add(nounPhrase.trim() );
+ nounPhrase = "";
+ }
+ }
+ return words;
+ }
+
+ private void removeStopWords(List<String> words){
+ words.removeAll(stopWords);
+ }
+
+ private List<String> getStemmedWords(List<String> words) {
+ List<String> stemmedWords = new ArrayList<String>();
+ Morphology morpho = new Morphology();
+ for (String w : words) {
+ if(!(w.indexOf(" ") > 0)){
+ stemmedWords.add(morpho.stem(w));
+ } else {
+ stemmedWords.add(w);
+ }
+ }
+ return stemmedWords;
+ }
+
+}
Modified: trunk/sparql-query-generator/src/main/java/org/dllearner/sparqlquerygenerator/util/QuestionBasedStatementFilter.java
===================================================================
--- trunk/sparql-query-generator/src/main/java/org/dllearner/sparqlquerygenerator/util/QuestionBasedStatementFilter.java 2011-02-25 11:23:58 UTC (rev 2664)
+++ trunk/sparql-query-generator/src/main/java/org/dllearner/sparqlquerygenerator/util/QuestionBasedStatementFilter.java 2011-02-26 11:49:29 UTC (rev 2665)
@@ -1,5 +1,7 @@
package org.dllearner.sparqlquerygenerator.util;
+import java.util.Set;
+
import uk.ac.shef.wit.simmetrics.similaritymetrics.AbstractStringMetric;
import uk.ac.shef.wit.simmetrics.similaritymetrics.QGramsDistance;
@@ -11,12 +13,13 @@
public class QuestionBasedStatementFilter implements Selector {
- private String question;
+ private Set<String> questionWords;
private AbstractStringMetric metric;
private double threshold = 0.7;
- public QuestionBasedStatementFilter(String question){
- this.question = question;
+
+ public QuestionBasedStatementFilter(Set<String> questionWords){
+ this.questionWords = questionWords;
metric = new QGramsDistance();
}
@@ -24,17 +27,29 @@
@Override
public boolean test(Statement s) {
String predicate = s.getPredicate().getURI().substring(s.getPredicate().getURI().lastIndexOf("/"));
- String object;
+ String object = null;
if(s.getObject().isURIResource()){
object = s.getObject().asResource().getURI();
object = object.substring(object.lastIndexOf("/"));
} else if(s.getObject().isLiteral()){
object = s.getObject().asLiteral().getLexicalForm();
}
+ if(isSimiliar2QuestionWord(object) || isSimiliar2QuestionWord(predicate)){
+ return true;
+ }
return false;
}
+ private boolean isSimiliar2QuestionWord(String s){
+ for(String word : questionWords){
+ if(areSimiliar(word, s)){
+ return true;
+ }
+ }
+ return false;
+ }
+
private boolean areSimiliar(String s1, String s2){
float sim = metric.getSimilarity(s1, s2);
return sim >= threshold;
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
