Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner:[1767] trunk/src/dl-learner/org/dllearner
|
From: <jen...@us...> - 2009-05-25 16:15:44
|
Revision: 1767
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=1767&view=rev
Author: jenslehmann
Date: 2009-05-25 16:15:37 +0000 (Mon, 25 May 2009)
Log Message:
-----------
extended DBpedia->LGD matching to cover 7 different POI types
Modified Paths:
--------------
trunk/src/dl-learner/org/dllearner/kb/sparql/SPARQLTasks.java
trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaLinkedGeoData.java
trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaPoint.java
trunk/src/dl-learner/org/dllearner/scripts/matching/Evaluation.java
trunk/src/dl-learner/org/dllearner/scripts/matching/LGDPoint.java
trunk/src/dl-learner/org/dllearner/scripts/matching/POIClass.java
Modified: trunk/src/dl-learner/org/dllearner/kb/sparql/SPARQLTasks.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/kb/sparql/SPARQLTasks.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/kb/sparql/SPARQLTasks.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -52,7 +52,7 @@
* the Endpoint the sparql queries will be send to
*/
public SPARQLTasks(final SparqlEndpoint sparqlEndpoint) {
- super();
+// super();
this.cache = null;
this.sparqlEndpoint = sparqlEndpoint;
}
@@ -64,7 +64,7 @@
* the Endpoint the sparql queries will be send to
*/
public SPARQLTasks(final Cache cache, final SparqlEndpoint sparqlEndpoint) {
- super();
+// super();
this.cache = cache;
this.sparqlEndpoint = sparqlEndpoint;
}
@@ -472,7 +472,6 @@
/**
* low level, executes query returns ResultSet.
- * TODO: Why convert from result set to JSON and back? See method below.
*
* @param sparqlQueryString
* The query
@@ -480,7 +479,13 @@
*/
public ResultSetRewindable queryAsResultSet(String sparqlQueryString) {
SparqlQuery sq = new SparqlQuery(sparqlQueryString, sparqlEndpoint);
- return sq.send();
+ if(cache == null) {
+ return sq.send();
+ } else {
+ // get JSON from cache and convert to result set
+ String json = cache.executeSparqlQuery(sq);
+ return SparqlQuery.convertJSONtoResultSet(json);
+ }
}
/**
Modified: trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaLinkedGeoData.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaLinkedGeoData.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaLinkedGeoData.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -70,7 +70,7 @@
// public static SparqlEndpoint dbpediaEndpoint = SparqlEndpoint.getEndpointDBpedia();
public static SparqlEndpoint dbpediaEndpoint = SparqlEndpoint.getEndpointLOCALDBpedia();
private static SparqlEndpoint geoDataEndpoint = SparqlEndpoint.getEndpointLOCALGeoData();
- private static SPARQLTasks lgd = new SPARQLTasks(new Cache("cache/matcher/"), geoDataEndpoint);
+ private static SPARQLTasks lgd = new SPARQLTasks(new Cache("cache/lgd/"), geoDataEndpoint);
// read in DBpedia ontology such that we perform taxonomy reasoning
// private static ReasonerComponent reasoner = TestOntologies.getTestOntology(TestOntology.DBPEDIA_OWL);
@@ -266,27 +266,39 @@
fos.close();
}
+ /**
+ * The main matching method. The matching is directed from DBpedia to LGD,
+ * i.e. given a POI in DBpedia, we try to find a match in LGD.
+ *
+ * @param dbpediaPoint The DBpedia point.
+ * @return The URI of the matched LGD point or null if no match was found.
+ * @throws IOException Thrown if a query or linked data access does not work.
+ */
public static URI findGeoDataMatch(DBpediaPoint dbpediaPoint) throws IOException {
// 1 degree is about 111 km (depending on the specific point)
- int distanceThresholdMeters = 1000;
+ double distanceThresholdMeters = dbpediaPoint.getPoiClass().getMaxBox();
boolean quiet = true;
if(useSparqlForGettingNearbyPoints) {
- // TODO: convert from meters to lat/long
- double distanceThresholdLat = 0.5;
- double distanceThresholdLong = 0.5;
-
- // Triplify: $1= , $2= , $3 = distance in meters
+ // deprecated: direct specification of long/lat difference
+// double distanceThresholdLat = 0.5;
+// double distanceThresholdLong = 0.5;
+ // create a box around the point
+// double minLat2 = dbpediaPoint.getGeoLat() - distanceThresholdLat;
+// double maxLat2 = dbpediaPoint.getGeoLat() + distanceThresholdLat;
+// double minLong2 = dbpediaPoint.getGeoLong() - distanceThresholdLong;
+// double maxLong2 = dbpediaPoint.getGeoLong() + distanceThresholdLong;
+
+ // Triplify: $1 = latitude, $2 = longitude, $3 = distance in meters
+ // LGD uses integer for lat/long (standard values multiplied by 10000000)
// $box='longitude between CEIL(($2-($3/1000)/abs(cos(radians($1))*111))*10000000) and CEIL(($2+($3/1000)/abs(cos(radians($1))*111))*10000000)
// AND latitude between CEIL(($1-($3/1000/111))*10000000) and CEIL(($1+($3/1000/111))*10000000)';
- // create a box around the point
- double minLat = dbpediaPoint.getGeoLat() - distanceThresholdLat;
- double maxLat = dbpediaPoint.getGeoLat() + distanceThresholdLat;
- double minLong = dbpediaPoint.getGeoLong() - distanceThresholdLong;
- double maxLong = dbpediaPoint.getGeoLong() + distanceThresholdLong;
-
+ double minLat = dbpediaPoint.getGeoLat()-(distanceThresholdMeters/1000/111);
+ double maxLat = dbpediaPoint.getGeoLat()+(distanceThresholdMeters/1000/111);
+ double minLong = dbpediaPoint.getGeoLong()-(distanceThresholdMeters/1000)/Math.abs(Math.cos(Math.toRadians(dbpediaPoint.getGeoLat()))*111);
+ double maxLong = dbpediaPoint.getGeoLong()+(distanceThresholdMeters/1000)/Math.abs(Math.cos(Math.toRadians(dbpediaPoint.getGeoLat()))*111);
// query all points in the box corresponding to this class
// (we make sure that returned points are in the same POI class)
String queryStr = "select ?point ?lat ?long ?name ?name_en ?name_int where { ";
Modified: trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaPoint.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaPoint.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/scripts/matching/DBpediaPoint.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -68,7 +68,7 @@
queryStr += "<"+uri+"> rdfs:label ?label . ";
queryStr += "OPTIONAL { <"+uri+"> rdf:type ?type . ";
queryStr += "FILTER (!(?type LIKE <http://dbpedia.org/ontology/Resource>)) .";
- queryStr += "FILTER (?type LIKE <http://dbpedia.org/ontology/%>) .";
+ queryStr += "FILTER (?type LIKE <http://dbpedia.org/ontology/%> || ?type LIKE <http://umbel.org/umbel/sc/%>) .";
queryStr += "} }";
// SparqlQuery query = new SparqlQuery(queryStr, DBpediaLinkedGeoData.dbpediaEndpoint);
@@ -147,10 +147,28 @@
return str + ")";
}
+ // maps classes (in DBpedia ontology or otherwise) to supported POI classes
private POIClass getPOIClass(String[] classes) {
for(String clazz : classes) {
+// System.out.println("class: " + clazz);
if(clazz.equals("http://dbpedia.org/ontology/City")) {
return POIClass.CITY;
+ } if(clazz.equals("http://umbel.org/umbel/sc/City")) {
+ return POIClass.CITY;
+ } else if(clazz.equals("http://dbpedia.org/ontology/Lake")) {
+ return POIClass.LAKE;
+ } else if(clazz.equals("http://dbpedia.org/ontology/University")) {
+ return POIClass.UNIVERSITY;
+ } else if(clazz.equals("http://dbpedia.org/ontology/School")) {
+ return POIClass.SCHOOL;
+ } else if(clazz.equals("http://dbpedia.org/ontology/Country")) {
+ return POIClass.COUNTRY;
+ } else if(clazz.equals("http://dbpedia.org/ontology/Airport")) {
+ return POIClass.AIRPORT;
+ } else if(clazz.equals("http://umbel.org/umbel/sc/Airfield")) {
+ return POIClass.AIRPORT;
+ } else if(clazz.equals("http://dbpedia.org/ontology/Station")) {
+ return POIClass.RAILWAY_STATION;
}
}
return null;
Modified: trunk/src/dl-learner/org/dllearner/scripts/matching/Evaluation.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/matching/Evaluation.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/scripts/matching/Evaluation.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -22,6 +22,8 @@
import java.io.File;
import java.io.IOException;
import java.net.URI;
+import java.util.Date;
+import java.util.HashMap;
import java.util.Map;
import java.util.Map.Entry;
import java.util.zip.DataFormatException;
@@ -47,6 +49,11 @@
private double precision;
private double recall;
+ private Map<POIClass, Integer> testsPerClass = new HashMap<POIClass, Integer>();
+ private Map<POIClass, Integer> noMatchPerClass = new HashMap<POIClass, Integer>();
+ private Map<POIClass, Integer> correctMatchPerClass = new HashMap<POIClass, Integer>();
+ private Map<POIClass, Integer> incorrectMatchPerClass = new HashMap<POIClass, Integer>();
+
private static Logger logger = Logger.getLogger(Evaluation.class);
// map from DBpedia to LinkedGeoData
@@ -58,6 +65,14 @@
correctMatchCount = 0;
incorrectMatchCount = 0;
+ // init counts
+ for(POIClass poiClass : POIClass.values()) {
+ testsPerClass.put(poiClass, 0);
+ noMatchPerClass.put(poiClass, 0);
+ correctMatchPerClass.put(poiClass, 0);
+ incorrectMatchPerClass.put(poiClass, 0);
+ }
+
for(Entry<URI,URI> match : testMatches.entrySet()) {
// find point in DBpedia file:
// approach 1:
@@ -82,12 +97,16 @@
URI matchedURI = null;
- if(dbpediaPoint.getPoiClass() == POIClass.CITY) {
- logger.info("Eval: searching match for " + match.getKey() + "...");
+ if(dbpediaPoint.getPoiClass() == null) {
+ if(dbpediaPoint.getClasses().length == 0) {
+ System.out.println("skipping " + dbpediaPoint.getUri() + " (unknown POI type)");
+ } else {
+ System.out.println("skipping " + dbpediaPoint.getUri() + " (unsupported POI type)");
+ }
+ continue;
+ } else {
+ logger.info("Eval: searching match for " + match.getKey() + "(" + dbpediaPoint.getPoiClass() + ") ...");
matchedURI = DBpediaLinkedGeoData.findGeoDataMatch(dbpediaPoint);
- } else {
- System.out.println("skipping " + dbpediaPoint.getUri() + " (not detected as a city)");
- continue;
}
URI testURI = match.getValue();
@@ -95,18 +114,22 @@
// no match found
if(matchedURI == null) {
noMatchCount++;
+ inc(noMatchPerClass, dbpediaPoint.getPoiClass());
logger.info("Eval: ... no match found");
// correct match found
} else if(matchedURI.equals(testURI)) {
correctMatchCount++;
+ inc(correctMatchPerClass, dbpediaPoint.getPoiClass());
logger.info("Eval: ... " + testURI + " correctly detected");
// incorrect match found
} else {
incorrectMatchCount++;
+ inc(incorrectMatchPerClass, dbpediaPoint.getPoiClass());
logger.info("Eval: ... " + matchedURI + " detected, but " + testURI + " is correct");
}
tests++;
+ inc(testsPerClass, dbpediaPoint.getPoiClass());
}
matchCount = correctMatchCount + incorrectMatchCount;
@@ -148,6 +171,50 @@
return discarded;
}
+ private void inc(Map<POIClass,Integer> map, POIClass poiClass) {
+// if(map.containsKey(poiClass)) {
+ map.put(poiClass, map.get(poiClass)+1);
+// } else {
+// map.put(poiClass, 1);
+// }
+ }
+
+ public Integer getCorrectMatchPerClass(POIClass poiClass) {
+ return correctMatchPerClass.get(poiClass);
+ }
+
+ public Integer getIncorrectMatchPerClass(POIClass poiClass) {
+ return incorrectMatchPerClass.get(poiClass);
+ }
+
+ public Integer getTestsPerClass(POIClass poiClass) {
+ return testsPerClass.get(poiClass);
+ }
+
+ public Integer getMatchPerClass(POIClass poiClass) {
+ return incorrectMatchPerClass.get(poiClass) + correctMatchPerClass.get(poiClass);
+ }
+
+ public Integer getNoMatchPerClass(POIClass poiClass) {
+ return noMatchPerClass.get(poiClass);
+ }
+
+ public double getPrecisionPerClass(POIClass poiClass) {
+ if(getMatchPerClass(poiClass) == 0) {
+ return 0;
+ } else {
+ return correctMatchPerClass.get(poiClass) / (double) getMatchPerClass(poiClass);
+ }
+ }
+
+ public double getRecallPerClass(POIClass poiClass) {
+ if(testsPerClass.get(poiClass) == 0) {
+ return 0;
+ } else {
+ return correctMatchPerClass.get(poiClass) / (double) testsPerClass.get(poiClass);
+ }
+ }
+
public static void main(String args[]) throws IOException, DataFormatException {
Logger.getRootLogger().setLevel(Level.INFO);
@@ -156,10 +223,24 @@
// map for collecting matches
Map<URI,URI> matches = Utility.getMatches(testFile);
// perform evaluation and print results
+ System.out.println(new Date());
Evaluation eval = new Evaluation(matches);
+ System.out.println(new Date());
+
+ for(POIClass poiClass : POIClass.values()) {
+ System.out.println();
+ System.out.println("summary for POI class " + poiClass + ":");
+ System.out.println(eval.getTestsPerClass(poiClass) + " points tested");
+ System.out.println("precision: " + eval.getPrecisionPerClass(poiClass) + " (" + eval.getCorrectMatchPerClass(poiClass) + "/" + eval.getMatchPerClass(poiClass) + ")");
+ System.out.println("recall: " + eval.getRecallPerClass(poiClass) + " (" + eval.getCorrectMatchPerClass(poiClass) + "/" + eval.getTestsPerClass(poiClass) + ")");
+ }
+
+ System.out.println("");
+ System.out.println("Overall summary:");
System.out.println(eval.getTests() + " points tested (" + eval.getDiscarded() + " discarded)");
System.out.println("precision: " + eval.getPrecision() + " (" + eval.getCorrectMatchCount() + "/" + eval.getMatchCount() + ")");
System.out.println("recall: " + eval.getRecall() + " (" + eval.getCorrectMatchCount() + "/" + eval.getTests() + ")");
+
}
}
Modified: trunk/src/dl-learner/org/dllearner/scripts/matching/LGDPoint.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/matching/LGDPoint.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/scripts/matching/LGDPoint.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -52,6 +52,12 @@
public static String getSPARQLRestriction(POIClass poiClass, String variable) {
switch(poiClass) {
case CITY : return "{ " + variable + " <http://linkedgeodata.org/vocabulary#place> \"city\" } UNION {" + variable + " <http://linkedgeodata.org/vocabulary#place> \"village\" } UNION {" + variable + " <http://linkedgeodata.org/vocabulary#place> \"town\" } UNION {" + variable + " <http://linkedgeodata.org/vocabulary#place> \"suburb\" }";
+ case UNIVERSITY : return variable + " <http://linkedgeodata.org/vocabulary#amenity> \"university\" . ";
+ case SCHOOL : return variable + " <http://linkedgeodata.org/vocabulary#amenity> \"school\" . ";
+ case AIRPORT : return variable + " <http://linkedgeodata.org/vocabulary#aeroway> \"aerodrome\" . ";
+ case LAKE : return variable + " <http://linkedgeodata.org/vocabulary#natural> \"water\" . ";
+ case COUNTRY : return variable + " <http://linkedgeodata.org/vocabulary#place> \"country\" . ";
+ case RAILWAY_STATION : return variable + " <http://linkedgeodata.org/vocabulary#railway> \"station\" . ";
default: throw new Error("Cannot restrict.");
}
}
Modified: trunk/src/dl-learner/org/dllearner/scripts/matching/POIClass.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/scripts/matching/POIClass.java 2009-05-22 13:47:11 UTC (rev 1766)
+++ trunk/src/dl-learner/org/dllearner/scripts/matching/POIClass.java 2009-05-25 16:15:37 UTC (rev 1767)
@@ -28,6 +28,44 @@
*/
public enum POIClass {
- CITY, COUNTRY
+ // 50 km box
+ CITY (50000),
+ // 5 km box
+ AIRPORT (5000),
+
+ // 10 km box
+ UNIVERSITY (10000),
+
+ // 10 km box - usage unclear
+// MUNICIPALITY (10000),
+
+ SCHOOL (10000),
+
+ RAILWAY_STATION (10000),
+
+ // 1000 km box
+ LAKE (1000000000),
+
+ // 5000 km box
+ COUNTRY (5000000);
+
+ private double maxBox;
+
+ POIClass(double maxBox) {
+ this.maxBox = maxBox;
+ }
+
+ /**
+ * Maximum distance coordinates and actual position of
+ * this POI type can differ in meters. Retrieving POIs within
+ * the box specified by this distance should always contain
+ * the POI itself.
+ *
+ * @return The distance in meters different typical POIs for this
+ * type can differ.
+ */
+ public double getMaxBox() {
+ return maxBox;
+ }
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
