public static void test() throws Exception {
ServletContextParameterMap contextParameters =
ContextParametersRegistry.getInstance().getDefault();
ModelingConfiguration modelingConfiguration =
ModelingConfigurationRegistry.getInstance()
.getModelingConfiguration(contextParameters.getId());
// String inputPath = Params.INPUT_DIR;
String outputPath = Params.OUTPUT_DIR;
String graphPath = Params.GRAPHS_DIR;
// List<SemanticModel> semanticModels = ModelReader.importSemanticModels(inputPath);
List<SemanticModel> semanticModels =
ModelReader.importSemanticModelsFromJsonFiles(Params.MODEL_DIR, Params.MODEL_MAIN_FILE_EXT);
// ModelEvaluation me2 = semanticModels.get(20).evaluate(semanticModels.get(20));
// System.out.println(me2.getPrecision() + "--" + me2.getRecall());
// if (true)
// return;
List<SemanticModel> trainingData = new ArrayList<SemanticModel>();
OntologyManager ontologyManager = new OntologyManager(contextParameters.getId());
File ff = new File(Params.ONTOLOGY_DIR);
File[] files = ff.listFiles();
for (File f : files) {
ontologyManager.doImport(f, "UTF-8");
}
ontologyManager.updateCache();
// getStatistics1(semanticModels);
// if (true)
// return;
ModelLearningGraph modelLearningGraph = null;
ModelLearner_Old modelLearner;
boolean iterativeEvaluation = false;
boolean useCorrectType = false;
int numberOfCRFCandidates = 4;
int numberOfKnownModels;
String filePath = Params.RESULTS_DIR;
String filename = "results,k=" + numberOfCRFCandidates + ".csv";
PrintWriter resultFile = new PrintWriter(new File(filePath + filename));
StringBuffer[] resultsArray = new StringBuffer[semanticModels.size() + 2];
for (int i = 0; i < resultsArray.length; i++) {
resultsArray[i] = new StringBuffer();
}
for (int i = 0; i < semanticModels.size(); i++) {
// for (int i = 0; i <= 10; i++) {
// int i = 3; {
resultFile.flush();
int newSourceIndex = i;
SemanticModel newSource = semanticModels.get(newSourceIndex);
logger.info("======================================================");
logger.info(newSource.getName() + "(#attributes:" + newSource.getColumnNodes().size() + ")");
System.out.println(
newSource.getName() + "(#attributes:" + newSource.getColumnNodes().size() + ")");
logger.info("======================================================");
if (!iterativeEvaluation) numberOfKnownModels = semanticModels.size() - 1;
else numberOfKnownModels = 0;
if (resultsArray[0].length() > 0) resultsArray[0].append(" \t ");
resultsArray[0].append(
newSource.getName()
+ "("
+ newSource.getColumnNodes().size()
+ ")"
+ "\t"
+ " "
+ "\t"
+ " ");
if (resultsArray[1].length() > 0) resultsArray[1].append(" \t ");
resultsArray[1].append("p \t r \t t");
while (numberOfKnownModels <= semanticModels.size() - 1) {
trainingData.clear();
int j = 0, count = 0;
while (count < numberOfKnownModels) {
if (j != newSourceIndex) {
trainingData.add(semanticModels.get(j));
count++;
}
j++;
}
modelLearningGraph =
(ModelLearningGraphSparse)
ModelLearningGraph.getEmptyInstance(ontologyManager, ModelLearningGraphType.Sparse);
SemanticModel correctModel = newSource;
List<ColumnNode> columnNodes = correctModel.getColumnNodes();
// if (useCorrectType && numberOfCRFCandidates > 1)
// updateCrfSemanticTypesForResearchEvaluation(columnNodes);
modelLearner = new ModelLearner_Old(ontologyManager, columnNodes);
long start = System.currentTimeMillis();
String graphName =
!iterativeEvaluation
? graphPath + semanticModels.get(newSourceIndex).getName() + Params.GRAPH_FILE_EXT
: graphPath
+ semanticModels.get(newSourceIndex).getName()
+ ".knownModels="
+ numberOfKnownModels
+ Params.GRAPH_FILE_EXT;
if (new File(graphName).exists()) {
// read graph from file
try {
logger.info("loading the graph ...");
DirectedWeightedMultigraph<Node, DefaultLink> graph = GraphUtil.importJson(graphName);
modelLearner.graphBuilder = new GraphBuilder(ontologyManager, graph, false);
modelLearner.nodeIdFactory = modelLearner.graphBuilder.getNodeIdFactory();
} catch (Exception e) {
e.printStackTrace();
}
} else {
logger.info("building the graph ...");
for (SemanticModel sm : trainingData) modelLearningGraph.addModel(sm, false);
modelLearner.graphBuilder = modelLearningGraph.getGraphBuilder();
modelLearner.nodeIdFactory = modelLearner.graphBuilder.getNodeIdFactory();
// save graph to file
try {
GraphUtil.exportJson(
modelLearningGraph.getGraphBuilder().getGraph(), graphName, true, true);
} catch (Exception e) {
e.printStackTrace();
}
}
List<SortableSemanticModel_Old> hypothesisList =
modelLearner.hypothesize(useCorrectType, numberOfCRFCandidates);
long elapsedTimeMillis = System.currentTimeMillis() - start;
float elapsedTimeSec = elapsedTimeMillis / 1000F;
List<SortableSemanticModel_Old> topHypotheses = null;
if (hypothesisList != null) {
topHypotheses =
hypothesisList.size() > modelingConfiguration.getNumCandidateMappings()
? hypothesisList.subList(0, modelingConfiguration.getNumCandidateMappings())
: hypothesisList;
}
Map<String, SemanticModel> models = new TreeMap<String, SemanticModel>();
// export to json
// if (topHypotheses != null)
// for (int k = 0; k < topHypotheses.size() && k < 3; k++) {
//
// String fileExt = null;
// if (k == 0) fileExt = Params.MODEL_RANK1_FILE_EXT;
// else if (k == 1) fileExt = Params.MODEL_RANK2_FILE_EXT;
// else if (k == 2) fileExt = Params.MODEL_RANK3_FILE_EXT;
// SortableSemanticModel m = topHypotheses.get(k);
// new SemanticModel(m).writeJson(Params.MODEL_DIR +
// newSource.getName() + fileExt);
//
// }
ModelEvaluation me;
models.put("1-correct model", correctModel);
if (topHypotheses != null)
for (int k = 0; k < topHypotheses.size(); k++) {
SortableSemanticModel_Old m = topHypotheses.get(k);
me = m.evaluate(correctModel);
String label =
"candidate"
+ k
+ m.getSteinerNodes().getScoreDetailsString()
+ "cost:"
+ roundTwoDecimals(m.getCost())
+
// "-distance:" + me.getDistance() +
"-precision:"
+ me.getPrecision()
+ "-recall:"
+ me.getRecall();
models.put(label, m);
if (k == 0) { // first rank model
System.out.println(
"number of known models: "
+ numberOfKnownModels
+ ", precision: "
+ me.getPrecision()
+ ", recall: "
+ me.getRecall()
+ ", time: "
+ elapsedTimeSec);
logger.info(
"number of known models: "
+ numberOfKnownModels
+ ", precision: "
+ me.getPrecision()
+ ", recall: "
+ me.getRecall()
+ ", time: "
+ elapsedTimeSec);
// resultFile.println("number of known models \t precision \t recall");
// resultFile.println(numberOfKnownModels + "\t" + me.getPrecision() + "\t" +
// me.getRecall());
String s = me.getPrecision() + "\t" + me.getRecall() + "\t" + elapsedTimeSec;
if (resultsArray[numberOfKnownModels + 2].length() > 0)
resultsArray[numberOfKnownModels + 2].append(" \t ");
resultsArray[numberOfKnownModels + 2].append(s);
// resultFile.println(me.getPrecision() + "\t" + me.getRecall() + "\t" +
// elapsedTimeSec);
}
}
String outName =
!iterativeEvaluation
? outputPath
+ semanticModels.get(newSourceIndex).getName()
+ Params.GRAPHVIS_OUT_DETAILS_FILE_EXT
: outputPath
+ semanticModels.get(newSourceIndex).getName()
+ ".knownModels="
+ numberOfKnownModels
+ Params.GRAPHVIS_OUT_DETAILS_FILE_EXT;
// if (!iterativeEvaluation) {
GraphVizUtil.exportSemanticModelsToGraphviz(
models,
newSource.getName(),
outName,
GraphVizLabelType.LocalId,
GraphVizLabelType.LocalUri,
false,
false);
// }
numberOfKnownModels++;
}
// resultFile.println("=======================================================");
}
for (StringBuffer s : resultsArray) resultFile.println(s.toString());
resultFile.close();
}
public List<SortableSemanticModel_Old> hypothesize(
boolean useCorrectTypes, int numberOfCRFCandidates) {
Set<Node> addedNodes =
new HashSet<
Node>(); // They should be deleted from the graph after computing the semantic models
logger.info("finding candidate steiner sets ... ");
CandidateSteinerSets candidateSteinerSets =
getCandidateSteinerSets(columnNodes, useCorrectTypes, numberOfCRFCandidates, addedNodes);
if (candidateSteinerSets == null
|| candidateSteinerSets.getSteinerSets() == null
|| candidateSteinerSets.getSteinerSets().isEmpty()) {
logger.error("there is no candidate set of steiner nodes.");
return null;
}
logger.info("number of steiner sets: " + candidateSteinerSets.numberOfCandidateSets());
logger.info("updating weights according to training data ...");
long start = System.currentTimeMillis();
this.updateWeights();
long updateWightsElapsedTimeMillis = System.currentTimeMillis() - start;
logger.info("time to update weights: " + (updateWightsElapsedTimeMillis / 1000F));
logger.info("computing steiner trees ...");
List<SortableSemanticModel_Old> sortableSemanticModels =
new ArrayList<SortableSemanticModel_Old>();
int count = 1;
for (SteinerNodes sn : candidateSteinerSets.getSteinerSets()) {
logger.debug("computing steiner tree for steiner nodes set " + count + " ...");
logger.debug(sn.getScoreDetailsString());
DirectedWeightedMultigraph<Node, LabeledLink> tree = computeSteinerTree(sn.getNodes());
count++;
if (tree != null) {
SemanticModel sm =
new SemanticModel(
new RandomGUID().toString(), tree, columnNodes, sn.getMappingToSourceColumns());
SortableSemanticModel_Old sortableSemanticModel = new SortableSemanticModel_Old(sm, sn);
sortableSemanticModels.add(sortableSemanticModel);
}
if (count
== ModelingConfigurationRegistry.getInstance()
.getModelingConfiguration(
ContextParametersRegistry.getInstance()
.getContextParameters(ontologyManager.getContextId())
.getKarmaHome())
.getNumCandidateMappings()) break;
}
Collections.sort(sortableSemanticModels);
// logger.info("results are ready ...");
// return sortableSemanticModels;
List<SortableSemanticModel_Old> uniqueModels = new ArrayList<SortableSemanticModel_Old>();
SortableSemanticModel_Old current, previous;
if (sortableSemanticModels != null) {
if (sortableSemanticModels.size() > 0) uniqueModels.add(sortableSemanticModels.get(0));
for (int i = 1; i < sortableSemanticModels.size(); i++) {
current = sortableSemanticModels.get(i);
previous = sortableSemanticModels.get(i - 1);
if (current.getScore() == previous.getScore() && current.getCost() == previous.getCost())
continue;
uniqueModels.add(current);
}
}
logger.info("results are ready ...");
return uniqueModels;
}
