/**
* runs the part of speech tagger
*
* @param file input file
* @param lang language
* @param type source or target
* @return path to the output file of the POS tagger
*/
public String runPOS(String file, String lang, String type) {
String posName = resourceManager.getString(lang + ".postagger");
String langResPath = input + File.separator + lang;
File f = new File(file);
String absoluteSourceFilePath = f.getAbsolutePath();
String fileName = f.getName();
String relativeFilePath = langResPath + File.separator + fileName + ".pos";
String absoluteOutputFilePath = (new File(relativeFilePath)).getAbsolutePath();
String posSourceTaggerPath = resourceManager.getString(lang + ".postagger.exePath");
String outPath = "";
try {
Class c = Class.forName(posName);
PosTagger tagger = (PosTagger) c.newInstance();
tagger.setParameters(
type, posName, posSourceTaggerPath, absoluteSourceFilePath, absoluteOutputFilePath);
PosTagger.ForceRun(forceRun);
outPath = tagger.run();
} catch (Exception e) {
e.printStackTrace();
}
// returns the path of the output file; this is for convenience only so
// we do't have to calculate it again
return outPath;
}
/** runs the BB features */
public void runBB() {
File f = new File(sourceFile);
String sourceFileName = f.getName();
f = new File(targetFile);
String targetFileName = f.getName();
String outputFileName = sourceFileName + "_to_" + targetFileName + ".out";
File file = new File(resourceManager.getString("output"));
if (!file.exists()) {
System.err.println("Creating dir: " + resourceManager.getString("output"));
Logger.log("Creating dir: " + resourceManager.getString("output"));
file.mkdirs();
} else {
Logger.log("output dir exists: " + resourceManager.getString("output"));
}
String out = resourceManager.getString("output") + File.separator + outputFileName;
System.out.println("Output will be: " + out);
String pplSourcePath =
resourceManager.getString("input")
+ File.separator
+ sourceLang
+ File.separator
+ sourceFileName
+ resourceManager.getString("tools.ngram.output.ext");
String pplTargetPath =
resourceManager.getString("input")
+ File.separator
+ targetLang
+ File.separator
+ targetFileName
+ resourceManager.getString("tools.ngram.output.ext");
String pplPOSTargetPath =
resourceManager.getString("input")
+ File.separator
+ targetLang
+ File.separator
+ targetFileName
+ PosTagger.getXPOS()
+ resourceManager.getString("tools.ngram.output.ext");
runNGramPPL();
FileModel fm = new FileModel(sourceFile, resourceManager.getString(sourceLang + ".corpus"));
// FileModel fm = new FileModel(sourceFile,
// resourceManager.getString("source" + ".corpus"));
PPLProcessor pplProcSource =
new PPLProcessor(pplSourcePath, new String[] {"logprob", "ppl", "ppl1"});
PPLProcessor pplProcTarget =
new PPLProcessor(pplTargetPath, new String[] {"logprob", "ppl", "ppl1"});
String sourcePosOutput = null;
String targetPosOutput = null;
PPLProcessor pplPosTarget = null;
if (!isBaseline) {
sourcePosOutput = runPOS(sourceFile, sourceLang, "source");
targetPosOutput = runPOS(targetFile, targetLang, "target");
String targetPPLPos = runNGramPPLPos(targetPosOutput + PosTagger.getXPOS());
System.out.println("---------TARGET PPLPOS: " + targetPPLPos);
pplPosTarget =
new PPLProcessor(targetPPLPos, new String[] {"poslogprob", "posppl", "posppl1"});
}
loadGiza();
processNGrams();
boolean gl = false;
String temp0 = resourceManager.getString("GL");
if (null != temp0 && temp0.equals("1")) {
gl = true;
}
if (gl) {
loadGlobalLexicon();
}
// Preparing the indices for IR_similarity_features
Lucene sourceLuc = null;
Lucene targetLuc = null;
if (featureManager.hasFeature("1700")) {
// The indices reside under lang_resources path
String lang_resources = workDir + File.separator + "lang_resources";
// Indices are saved under: luceneIndex folder
String source_lucene_path =
lang_resources + File.separator + sourceLang + File.separator + "luceneIndex";
// The corpus to index
String source_lucene_corpus = source_lucene_path + File.separator + sourceLang + ".corpus";
// System.out.println("SOURCE: " + source_lucene_path + " ||| " + source_lucene_corpus);
try {
sourceLuc = new Lucene(source_lucene_path, source_lucene_corpus, true, true, "Source");
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
String target_lucene_path =
lang_resources + File.separator + targetLang + File.separator + "luceneIndex";
String target_lucene_corpus = target_lucene_path + File.separator + targetLang + ".corpus";
// System.out.println("TARGET: " + target_lucene_path + " ||| " + target_lucene_corpus);
try {
targetLuc = new Lucene(target_lucene_path, target_lucene_corpus, true, true, "Target");
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
// MQM kicks in
MQMManager.getInstance().initialize(resourceManager);
Context context = new Context();
context.setSourceFilePath(sourceFile);
context.setTargetFilePath(targetFile);
MQMManager.getInstance().globalProcessing(context);
try {
BufferedReader brSource = new BufferedReader(new FileReader(sourceFile));
BufferedReader brTarget = new BufferedReader(new FileReader(targetFile));
BufferedWriter output = new BufferedWriter(new FileWriter(out));
BufferedReader posSource = null;
BufferedReader posTarget = null;
boolean posSourceExists = ResourceManager.isRegistered("sourcePosTagger");
boolean posTargetExists = ResourceManager.isRegistered("targetPosTagger");
POSProcessor posSourceProc = null;
POSProcessor posTargetProc = null;
// lefterav: Berkeley parser modifications start here
// Check if user has defined the grammar files for source
// and target language
// if ( ResourceManager.isRegistered("BParser")){
boolean bp = false;
String temp = resourceManager.getString("BP");
if (null != temp && temp.equals("1")) {
bp = true;
}
BParserProcessor sourceParserProcessor = null;
BParserProcessor targetParserProcessor = null;
if (bp) {
sourceParserProcessor = new BParserProcessor();
targetParserProcessor = new BParserProcessor();
sourceParserProcessor.initialize(sourceFile, resourceManager, sourceLang);
targetParserProcessor.initialize(targetFile, resourceManager, targetLang);
}
// }
/** BEGIN: Added by Raphael Rubino for the Topic Model Features */
boolean tm = false;
String temp1 = resourceManager.getString("TM");
if (temp1 != null && temp1.equals("1")) {
tm = true;
}
TopicDistributionProcessor sourceTopicDistributionProcessor = null;
TopicDistributionProcessor targetTopicDistributionProcessor = null;
if (tm) {
String sourceTopicDistributionFile =
resourceManager.getString(sourceLang + ".topic.distribution");
String targetTopicDistributionFile =
resourceManager.getString(targetLang + ".topic.distribution");
sourceTopicDistributionProcessor =
new TopicDistributionProcessor(sourceTopicDistributionFile, "sourceTopicDistribution");
targetTopicDistributionProcessor =
new TopicDistributionProcessor(targetTopicDistributionFile, "targetTopicDistribution");
}
/* END: Added by Raphael Rubino for the Topic Model Features
*/
if (!isBaseline) {
if (posSourceExists) {
posSourceProc = new POSProcessor(sourcePosOutput);
posSource =
new BufferedReader(
new InputStreamReader(new FileInputStream(sourcePosOutput), "utf-8"));
}
if (posTargetExists) {
posTargetProc = new POSProcessor(targetPosOutput);
posTarget =
new BufferedReader(new InputStreamReader(new FileInputStream(targetPosOutput)));
}
}
ResourceManager.printResources();
Sentence sourceSent;
Sentence targetSent;
int sentCount = 0;
String lineSource = brSource.readLine();
String lineTarget = brTarget.readLine();
/** Triggers (by David Langlois) */
boolean tr = false;
String temp2 = resourceManager.getString("TR");
if (temp2 != null && temp2.equals("1")) {
tr = true;
}
Triggers itl_target = null;
TriggersProcessor itl_target_p = null;
Triggers itl_source = null;
TriggersProcessor itl_source_p = null;
// TriggersProcessor itl_source_p = null;
Triggers itl_source_target = null;
TriggersProcessor itl_source_target_p = null;
if (tr) {
itl_target =
new Triggers(
resourceManager.getString("target.intra.triggers.file"),
Integer.parseInt(resourceManager.getString("nb.max.triggers.target.intra")),
resourceManager.getString("phrase.separator"));
itl_target_p = new TriggersProcessor(itl_target);
itl_source =
new Triggers(
resourceManager.getString("source.intra.triggers.file"),
Integer.parseInt(resourceManager.getString("nb.max.triggers.source.intra")),
resourceManager.getString("phrase.separator"));
itl_source_p = new TriggersProcessor(itl_source);
itl_source_target =
new Triggers(
resourceManager.getString("source.target.inter.triggers.file"),
Integer.parseInt(resourceManager.getString("nb.max.triggers.source.target.inter")),
resourceManager.getString("phrase.separator"));
itl_source_target_p = new TriggersProcessor(itl_source_target);
}
/*
* End modification for Triggers
*/
// read in each line from the source and target files
// create a sentence from each
// process each sentence
// run the features on the sentences
while ((lineSource != null) && (lineTarget != null)) {
// lineSource = lineSource.trim().substring(lineSource.indexOf(" ")).replace("+", "");
sourceSent = new Sentence(lineSource, sentCount);
targetSent = new Sentence(lineTarget, sentCount);
// System.out.println("Processing sentence "+sentCount);
// System.out.println("SORCE: " + sourceSent.getText());
// System.out.println("TARGET: " + targetSent.getText());
if (posSourceExists) {
posSourceProc.processSentence(sourceSent);
}
if (posTargetExists) {
posTargetProc.processSentence(targetSent);
}
sourceSent.computeNGrams(3);
targetSent.computeNGrams(3);
pplProcSource.processNextSentence(sourceSent);
pplProcTarget.processNextSentence(targetSent);
if (!isBaseline) {
pplPosTarget.processNextSentence(targetSent);
}
// lefterav: Parse code here
if (bp) {
sourceParserProcessor.processNextSentence(sourceSent);
targetParserProcessor.processNextSentence(targetSent);
}
if (tm) {
sourceTopicDistributionProcessor.processNextSentence(sourceSent);
targetTopicDistributionProcessor.processNextSentence(targetSent);
}
// modified by David
if (tr) {
itl_source_p.processNextSentence(sourceSent);
itl_target_p.processNextSentence(targetSent);
itl_source_target_p.processNextParallelSentences(sourceSent, targetSent);
}
// end modification by David
// MQM kicks in
MQMManager.getInstance().processNextParallelSentences(sourceSent, targetSent);
// Ergun
if (featureManager.hasFeature("1700")) {
sourceLuc.processNextSentence(sourceSent);
targetLuc.processNextSentence(targetSent);
}
++sentCount;
output.write(featureManager.runFeatures(sourceSent, targetSent));
output.newLine();
lineSource = brSource.readLine();
lineTarget = brTarget.readLine();
}
if (posSource != null) {
posSource.close();
}
if (posTarget != null) {
posTarget.close();
}
brSource.close();
brTarget.close();
output.close();
Logger.close();
} catch (Exception e) {
e.printStackTrace();
}
}
public void runAll() {
File f = new File(sourceFile);
String sourceFileName = f.getName();
f = new File(targetFile);
String targetFileName = f.getName();
String outputFileName = sourceFileName + "_to_" + targetFileName + ".out";
String out = resourceManager.getString("output") + File.separator + outputFileName;
System.out.println("Output will be: " + out);
MTOutputProcessor mtop = null;
if (gbMode == 1) gbXML = initialiseGBResources();
String nbestSentPath =
resourceManager.getString("input") + File.separator + targetLang + File.separator + "temp";
String ngramExecPath = resourceManager.getString("tools.ngram.path");
mtop = new MTOutputProcessor(gbXML, nbestSentPath, ngramExecPath, ngramSize);
// wlv.mt.features.coherence.Coherence coh = new wlv.mt.features.coherence.Coherence(
// getTargetFile());
String pplSourcePath =
resourceManager.getString("input")
+ File.separator
+ sourceLang
+ File.separator
+ sourceFileName
+ resourceManager.getString("tools.ngram.output.ext");
String pplTargetPath =
resourceManager.getString("input")
+ File.separator
+ targetLang
+ File.separator
+ targetFileName
+ resourceManager.getString("tools.ngram.output.ext");
String pplPOSTargetPath =
resourceManager.getString("input")
+ File.separator
+ targetLang
+ File.separator
+ targetFileName
+ PosTagger.getXPOS()
+ resourceManager.getString("tools.ngram.output.ext");
runNGramPPL();
PPLProcessor pplProcSource =
new PPLProcessor(pplSourcePath, new String[] {"logprob", "ppl", "ppl1"});
PPLProcessor pplProcTarget =
new PPLProcessor(pplTargetPath, new String[] {"logprob", "ppl", "ppl1"});
FileModel fm = new FileModel(sourceFile, resourceManager.getString(sourceLang + ".corpus"));
String sourcePosOutput = runPOS(sourceFile, sourceLang, "source");
String targetPosOutput = runPOS(targetFile, targetLang, "target");
String targetPPLPos = runNGramPPLPos(targetPosOutput + PosTagger.getXPOS());
System.out.println("---------TARGET PPLPOS: " + targetPPLPos);
PPLProcessor pplPosTarget =
new PPLProcessor(targetPPLPos, new String[] {"poslogprob", "posppl", "posppl1"});
loadGiza();
processNGrams();
try {
BufferedReader brSource = new BufferedReader(new FileReader(sourceFileName));
BufferedReader brTarget = new BufferedReader(new FileReader(targetFileName));
BufferedWriter output = new BufferedWriter(new FileWriter(out));
BufferedReader posSource = null;
BufferedReader posTarget = null;
boolean posSourceExists = ResourceManager.isRegistered("sourcePosTagger");
boolean posTargetExists = ResourceManager.isRegistered("targetPosTagger");
POSProcessor posSourceProc = null;
POSProcessor posTargetProc = null;
if (posSourceExists) {
posSourceProc = new POSProcessor(sourcePosOutput);
posSource =
new BufferedReader(
new InputStreamReader(new FileInputStream(sourcePosOutput), "utf-8"));
}
if (posTargetExists) {
posTargetProc = new POSProcessor(targetPosOutput);
posTarget = new BufferedReader(new InputStreamReader(new FileInputStream(targetPosOutput)));
}
ResourceManager.printResources();
Sentence targetSent;
// HACK
Sentence sourceSent;
int sentCount = 0;
// HACK
String lineSource = brSource.readLine();
String lineTarget = brTarget.readLine();
// HACK
int result;
while ((lineSource != null) && (lineTarget != null)) {
// the MADA-tokenised files contain start each sentence with the setence ID. We put it there
// (why?) - no we've got to remove it
lineSource = lineSource.trim().substring(lineSource.indexOf(" ")).replace("+", "");
sourceSent = new Sentence(lineSource, sentCount);
targetSent = new Sentence(lineTarget, sentCount);
System.out.println("Processing sentence " + sentCount);
if (posSourceExists) {
posSourceProc.processSentence(sourceSent);
}
if (posTargetExists) {
posTargetProc.processSentence(targetSent);
}
sourceSent.computeNGrams(3);
targetSent.computeNGrams(3);
pplProcSource.processNextSentence(sourceSent);
pplProcTarget.processNextSentence(targetSent);
pplPosTarget.processNextSentence(targetSent);
// coh.processNextSentence(targetSent);
mtop.processNextSentence(sourceSent);
++sentCount;
output.write(featureManager.runFeatures(sourceSent, targetSent));
output.write("\r\n");
lineSource = brSource.readLine();
lineTarget = brTarget.readLine();
}
// featureManager.printFeatureIndeces();
if (posSource != null) {
posSource.close();
}
if (posTarget != null) {
posTarget.close();
}
brSource.close();
brTarget.close();
output.close();
Logger.close();
} catch (Exception e) {
e.printStackTrace();
}
}