/**
* 57% of time spent in LogConditionalObjectiveFunction.calculateCLBatch() 22% spent in
* constructing datums (expensive)
*
* <p>Single threaded, 4100 ms Multi threaded, 600 ms
*
* <p>With same data, seed 42, 52 ms With reordered accesses for cacheing, 38 ms Down to 73% of
* the time
*
* <p>with 8 cpus, a 6.8x speedup -- basically the same as with RVFDatum
*/
public static void benchmarkLogisticRegression() {
Dataset<String, String> data = new Dataset<>();
for (int i = 0; i < 10000; i++) {
Random r = new Random(42);
Set<String> features = new HashSet<>();
boolean cl = r.nextBoolean();
for (int j = 0; j < 1000; j++) {
if (cl && i % 2 == 0) {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":true");
} else {
features.add("f:" + j + ":false");
}
} else {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":false");
} else {
features.add("f:" + j + ":false");
}
}
}
data.add(new BasicDatum<String, String>(features, "target:" + cl));
}
LinearClassifierFactory<String, String> factory = new LinearClassifierFactory<>();
long msStart = System.currentTimeMillis();
factory.trainClassifier(data);
long delay = System.currentTimeMillis() - msStart;
System.out.println("Training took " + delay + " ms");
}
/**
* 67% of time spent in LogConditionalObjectiveFunction.rvfcalculate() 29% of time spent in
* dataset construction (11% in RVFDataset.addFeatures(), 7% rvf incrementCount(), 11% rest)
*
* <p>Single threaded, 4700 ms Multi threaded, 700 ms
*
* <p>With same data, seed 42, 245 ms With reordered accesses for cacheing, 195 ms Down to 80% of
* the time, not huge but a win nonetheless
*
* <p>with 8 cpus, a 6.7x speedup -- almost, but not quite linear, pretty good
*/
public static void benchmarkRVFLogisticRegression() {
RVFDataset<String, String> data = new RVFDataset<>();
for (int i = 0; i < 10000; i++) {
Random r = new Random(42);
Counter<String> features = new ClassicCounter<>();
boolean cl = r.nextBoolean();
for (int j = 0; j < 1000; j++) {
double value;
if (cl && i % 2 == 0) {
value = (r.nextDouble() * 2.0) - 0.6;
} else {
value = (r.nextDouble() * 2.0) - 1.4;
}
features.incrementCount("f" + j, value);
}
data.add(new RVFDatum<>(features, "target:" + cl));
}
LinearClassifierFactory<String, String> factory = new LinearClassifierFactory<>();
long msStart = System.currentTimeMillis();
factory.trainClassifier(data);
long delay = System.currentTimeMillis() - msStart;
System.out.println("Training took " + delay + " ms");
}
/** on my machine this results in a factor of two gain, roughly */
public static void testAdjacency() {
double[][] sqar = new double[10000][1000];
Random r = new Random();
int k = 0;
long msStart = System.currentTimeMillis();
for (int i = 0; i < 10000; i++) {
int loc = r.nextInt(10000);
for (int j = 0; j < 1000; j++) {
k += sqar[loc][j];
}
}
long delay = System.currentTimeMillis() - msStart;
System.out.println("Scanning with cache friendly lookups took " + delay + " ms");
int[] randLocs = new int[10000];
for (int i = 0; i < 10000; i++) {
randLocs[i] = r.nextInt(10000);
}
k = 0;
msStart = System.currentTimeMillis();
for (int j = 0; j < 1000; j++) {
for (int i = 0; i < 10000; i++) {
k += sqar[randLocs[i]][j];
}
}
delay = System.currentTimeMillis() - msStart;
System.out.println("Scanning with cache UNfriendly lookups took " + delay + " ms");
}
/**
* 29% in FactorTable.getValue() 28% in CRFCliqueTree.getCalibratedCliqueTree() 12.6% waiting for
* threads
*
* <p>Single threaded: 15000 ms - 26000 ms Multi threaded: 4500 ms - 7000 ms
*
* <p>with 8 cpus, 3.3x - 3.7x speedup, around 800% utilization
*/
public static void benchmarkCRF() {
Properties props = new Properties();
props.setProperty("macro", "true"); // use a generic CRF configuration
props.setProperty("useIfInteger", "true");
props.setProperty("featureFactory", "edu.stanford.nlp.benchmarks.BenchmarkFeatureFactory");
props.setProperty("saveFeatureIndexToDisk", "false");
CRFClassifier<CoreLabel> crf = new CRFClassifier<CoreLabel>(props);
Random r = new Random(42);
List<List<CoreLabel>> data = new ArrayList<>();
for (int i = 0; i < 100; i++) {
List<CoreLabel> sentence = new ArrayList<>();
for (int j = 0; j < 20; j++) {
CoreLabel l = new CoreLabel();
l.setWord("j:" + j);
boolean tag = j % 2 == 0 ^ (r.nextDouble() > 0.7);
l.set(CoreAnnotations.AnswerAnnotation.class, "target:" + tag);
sentence.add(l);
}
data.add(sentence);
}
long msStart = System.currentTimeMillis();
crf.train(data);
long delay = System.currentTimeMillis() - msStart;
System.out.println("Training took " + delay + " ms");
}
public static void benchmarkDatum() {
long msStart = System.currentTimeMillis();
Dataset<String, String> data = new Dataset<>();
for (int i = 0; i < 10000; i++) {
Random r = new Random(42);
Set<String> features = new HashSet<>();
boolean cl = r.nextBoolean();
for (int j = 0; j < 1000; j++) {
if (cl && i % 2 == 0) {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":true");
} else {
features.add("f:" + j + ":false");
}
} else {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":false");
} else {
features.add("f:" + j + ":false");
}
}
}
data.add(new BasicDatum<String, String>(features, "target:" + cl));
}
long delay = System.currentTimeMillis() - msStart;
System.out.println("Dataset construction took " + delay + " ms");
msStart = System.currentTimeMillis();
for (int i = 0; i < 10000; i++) {
Random r = new Random(42);
Set<String> features = new HashSet<>();
boolean cl = r.nextBoolean();
for (int j = 0; j < 1000; j++) {
if (cl && i % 2 == 0) {
if (r.nextDouble() > 0.3) {
} else {
}
} else {
if (r.nextDouble() > 0.3) {
} else {
}
}
}
}
delay = System.currentTimeMillis() - msStart;
System.out.println("MultiVector took " + delay + " ms");
}
public static void benchmarkSGD() {
Dataset<String, String> data = new Dataset<>();
for (int i = 0; i < 10000; i++) {
Random r = new Random(42);
Set<String> features = new HashSet<>();
boolean cl = r.nextBoolean();
for (int j = 0; j < 1000; j++) {
if (cl && i % 2 == 0) {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":true");
} else {
features.add("f:" + j + ":false");
}
} else {
if (r.nextDouble() > 0.3) {
features.add("f:" + j + ":false");
} else {
features.add("f:" + j + ":false");
}
}
}
data.add(new BasicDatum<String, String>(features, "target:" + cl));
}
LinearClassifierFactory<String, String> factory = new LinearClassifierFactory<>();
factory.setMinimizerCreator(
new Factory<Minimizer<DiffFunction>>() {
@Override
public Minimizer<DiffFunction> create() {
return new SGDMinimizer<DiffFunction>(0.1, 100, 0, 1000);
}
});
long msStart = System.currentTimeMillis();
factory.trainClassifier(data);
long delay = System.currentTimeMillis() - msStart;
System.out.println("Training took " + delay + " ms");
}
/**
* Prints out all matches of a semgrex pattern on a file of dependencies. <br>
* Usage:<br>
* java edu.stanford.nlp.semgraph.semgrex.SemgrexPattern [args] <br>
* See the help() function for a list of possible arguments to provide.
*/
public static void main(String[] args) throws IOException {
Map<String, Integer> flagMap = Generics.newHashMap();
flagMap.put(PATTERN, 1);
flagMap.put(TREE_FILE, 1);
flagMap.put(MODE, 1);
flagMap.put(EXTRAS, 1);
flagMap.put(CONLLU_FILE, 1);
flagMap.put(OUTPUT_FORMAT_OPTION, 1);
Map<String, String[]> argsMap = StringUtils.argsToMap(args, flagMap);
args = argsMap.get(null);
// TODO: allow patterns to be extracted from a file
if (!(argsMap.containsKey(PATTERN)) || argsMap.get(PATTERN).length == 0) {
help();
System.exit(2);
}
SemgrexPattern semgrex = SemgrexPattern.compile(argsMap.get(PATTERN)[0]);
String modeString = DEFAULT_MODE;
if (argsMap.containsKey(MODE) && argsMap.get(MODE).length > 0) {
modeString = argsMap.get(MODE)[0].toUpperCase();
}
SemanticGraphFactory.Mode mode = SemanticGraphFactory.Mode.valueOf(modeString);
String outputFormatString = DEFAULT_OUTPUT_FORMAT;
if (argsMap.containsKey(OUTPUT_FORMAT_OPTION) && argsMap.get(OUTPUT_FORMAT_OPTION).length > 0) {
outputFormatString = argsMap.get(OUTPUT_FORMAT_OPTION)[0].toUpperCase();
}
OutputFormat outputFormat = OutputFormat.valueOf(outputFormatString);
boolean useExtras = true;
if (argsMap.containsKey(EXTRAS) && argsMap.get(EXTRAS).length > 0) {
useExtras = Boolean.valueOf(argsMap.get(EXTRAS)[0]);
}
List<SemanticGraph> graphs = Generics.newArrayList();
// TODO: allow other sources of graphs, such as dependency files
if (argsMap.containsKey(TREE_FILE) && argsMap.get(TREE_FILE).length > 0) {
for (String treeFile : argsMap.get(TREE_FILE)) {
System.err.println("Loading file " + treeFile);
MemoryTreebank treebank = new MemoryTreebank(new TreeNormalizer());
treebank.loadPath(treeFile);
for (Tree tree : treebank) {
// TODO: allow other languages... this defaults to English
SemanticGraph graph =
SemanticGraphFactory.makeFromTree(
tree,
mode,
useExtras
? GrammaticalStructure.Extras.MAXIMAL
: GrammaticalStructure.Extras.NONE,
true);
graphs.add(graph);
}
}
}
if (argsMap.containsKey(CONLLU_FILE) && argsMap.get(CONLLU_FILE).length > 0) {
CoNLLUDocumentReader reader = new CoNLLUDocumentReader();
for (String conlluFile : argsMap.get(CONLLU_FILE)) {
System.err.println("Loading file " + conlluFile);
Iterator<SemanticGraph> it = reader.getIterator(IOUtils.readerFromString(conlluFile));
while (it.hasNext()) {
SemanticGraph graph = it.next();
graphs.add(graph);
}
}
}
for (SemanticGraph graph : graphs) {
SemgrexMatcher matcher = semgrex.matcher(graph);
if (!(matcher.find())) {
continue;
}
if (outputFormat == OutputFormat.LIST) {
System.err.println("Matched graph:");
System.err.println(graph.toString(SemanticGraph.OutputFormat.LIST));
boolean found = true;
while (found) {
System.err.println(
"Matches at: " + matcher.getMatch().value() + "-" + matcher.getMatch().index());
List<String> nodeNames = Generics.newArrayList();
nodeNames.addAll(matcher.getNodeNames());
Collections.sort(nodeNames);
for (String name : nodeNames) {
System.err.println(
" "
+ name
+ ": "
+ matcher.getNode(name).value()
+ "-"
+ matcher.getNode(name).index());
}
System.err.println();
found = matcher.find();
}
} else if (outputFormat == OutputFormat.OFFSET) {
if (graph.vertexListSorted().isEmpty()) {
continue;
}
System.out.printf(
"+%d %s%n",
graph.vertexListSorted().get(0).get(CoreAnnotations.LineNumberAnnotation.class),
argsMap.get(CONLLU_FILE)[0]);
}
}
}
public static void main(String[] args) {
Options op = new Options(new EnglishTreebankParserParams());
// op.tlpParams may be changed to something else later, so don't use it till
// after options are parsed.
System.out.println("Currently " + new Date());
System.out.print("Invoked with arguments:");
for (String arg : args) {
System.out.print(" " + arg);
}
System.out.println();
String path = "/u/nlp/stuff/corpora/Treebank3/parsed/mrg/wsj";
int trainLow = 200, trainHigh = 2199, testLow = 2200, testHigh = 2219;
String serializeFile = null;
int i = 0;
while (i < args.length && args[i].startsWith("-")) {
if (args[i].equalsIgnoreCase("-path") && (i + 1 < args.length)) {
path = args[i + 1];
i += 2;
} else if (args[i].equalsIgnoreCase("-train") && (i + 2 < args.length)) {
trainLow = Integer.parseInt(args[i + 1]);
trainHigh = Integer.parseInt(args[i + 2]);
i += 3;
} else if (args[i].equalsIgnoreCase("-test") && (i + 2 < args.length)) {
testLow = Integer.parseInt(args[i + 1]);
testHigh = Integer.parseInt(args[i + 2]);
i += 3;
} else if (args[i].equalsIgnoreCase("-serialize") && (i + 1 < args.length)) {
serializeFile = args[i + 1];
i += 2;
} else if (args[i].equalsIgnoreCase("-tLPP") && (i + 1 < args.length)) {
try {
op.tlpParams = (TreebankLangParserParams) Class.forName(args[i + 1]).newInstance();
} catch (ClassNotFoundException e) {
System.err.println("Class not found: " + args[i + 1]);
} catch (InstantiationException e) {
System.err.println("Couldn't instantiate: " + args[i + 1] + ": " + e.toString());
} catch (IllegalAccessException e) {
System.err.println("illegal access" + e);
}
i += 2;
} else if (args[i].equals("-encoding")) {
// sets encoding for TreebankLangParserParams
op.tlpParams.setInputEncoding(args[i + 1]);
op.tlpParams.setOutputEncoding(args[i + 1]);
i += 2;
} else {
i = op.setOptionOrWarn(args, i);
}
}
// System.out.println(tlpParams.getClass());
TreebankLanguagePack tlp = op.tlpParams.treebankLanguagePack();
Train.sisterSplitters = new HashSet(Arrays.asList(op.tlpParams.sisterSplitters()));
// BinarizerFactory.TreeAnnotator.setTreebankLang(tlpParams);
PrintWriter pw = op.tlpParams.pw();
Test.display();
Train.display();
op.display();
op.tlpParams.display();
// setup tree transforms
Treebank trainTreebank = op.tlpParams.memoryTreebank();
MemoryTreebank testTreebank = op.tlpParams.testMemoryTreebank();
// Treebank blippTreebank = ((EnglishTreebankParserParams) tlpParams).diskTreebank();
// String blippPath = "/afs/ir.stanford.edu/data/linguistic-data/BLLIP-WSJ/";
// blippTreebank.loadPath(blippPath, "", true);
Timing.startTime();
System.err.print("Reading trees...");
testTreebank.loadPath(path, new NumberRangeFileFilter(testLow, testHigh, true));
if (Test.increasingLength) {
Collections.sort(testTreebank, new TreeLengthComparator());
}
trainTreebank.loadPath(path, new NumberRangeFileFilter(trainLow, trainHigh, true));
Timing.tick("done.");
System.err.print("Binarizing trees...");
TreeAnnotatorAndBinarizer binarizer = null;
if (!Train.leftToRight) {
binarizer =
new TreeAnnotatorAndBinarizer(op.tlpParams, op.forceCNF, !Train.outsideFactor(), true);
} else {
binarizer =
new TreeAnnotatorAndBinarizer(
op.tlpParams.headFinder(),
new LeftHeadFinder(),
op.tlpParams,
op.forceCNF,
!Train.outsideFactor(),
true);
}
CollinsPuncTransformer collinsPuncTransformer = null;
if (Train.collinsPunc) {
collinsPuncTransformer = new CollinsPuncTransformer(tlp);
}
TreeTransformer debinarizer = new Debinarizer(op.forceCNF);
List<Tree> binaryTrainTrees = new ArrayList<Tree>();
if (Train.selectiveSplit) {
Train.splitters =
ParentAnnotationStats.getSplitCategories(
trainTreebank,
Train.tagSelectiveSplit,
0,
Train.selectiveSplitCutOff,
Train.tagSelectiveSplitCutOff,
op.tlpParams.treebankLanguagePack());
if (Train.deleteSplitters != null) {
List<String> deleted = new ArrayList<String>();
for (String del : Train.deleteSplitters) {
String baseDel = tlp.basicCategory(del);
boolean checkBasic = del.equals(baseDel);
for (Iterator<String> it = Train.splitters.iterator(); it.hasNext(); ) {
String elem = it.next();
String baseElem = tlp.basicCategory(elem);
boolean delStr = checkBasic && baseElem.equals(baseDel) || elem.equals(del);
if (delStr) {
it.remove();
deleted.add(elem);
}
}
}
System.err.println("Removed from vertical splitters: " + deleted);
}
}
if (Train.selectivePostSplit) {
TreeTransformer myTransformer = new TreeAnnotator(op.tlpParams.headFinder(), op.tlpParams);
Treebank annotatedTB = trainTreebank.transform(myTransformer);
Train.postSplitters =
ParentAnnotationStats.getSplitCategories(
annotatedTB,
true,
0,
Train.selectivePostSplitCutOff,
Train.tagSelectivePostSplitCutOff,
op.tlpParams.treebankLanguagePack());
}
if (Train.hSelSplit) {
binarizer.setDoSelectiveSplit(false);
for (Tree tree : trainTreebank) {
if (Train.collinsPunc) {
tree = collinsPuncTransformer.transformTree(tree);
}
// tree.pennPrint(tlpParams.pw());
tree = binarizer.transformTree(tree);
// binaryTrainTrees.add(tree);
}
binarizer.setDoSelectiveSplit(true);
}
for (Tree tree : trainTreebank) {
if (Train.collinsPunc) {
tree = collinsPuncTransformer.transformTree(tree);
}
tree = binarizer.transformTree(tree);
binaryTrainTrees.add(tree);
}
if (Test.verbose) {
binarizer.dumpStats();
}
List<Tree> binaryTestTrees = new ArrayList<Tree>();
for (Tree tree : testTreebank) {
if (Train.collinsPunc) {
tree = collinsPuncTransformer.transformTree(tree);
}
tree = binarizer.transformTree(tree);
binaryTestTrees.add(tree);
}
Timing.tick("done."); // binarization
BinaryGrammar bg = null;
UnaryGrammar ug = null;
DependencyGrammar dg = null;
// DependencyGrammar dgBLIPP = null;
Lexicon lex = null;
// extract grammars
Extractor bgExtractor = new BinaryGrammarExtractor();
// Extractor bgExtractor = new SmoothedBinaryGrammarExtractor();//new BinaryGrammarExtractor();
// Extractor lexExtractor = new LexiconExtractor();
// Extractor dgExtractor = new DependencyMemGrammarExtractor();
Extractor dgExtractor = new MLEDependencyGrammarExtractor(op);
if (op.doPCFG) {
System.err.print("Extracting PCFG...");
Pair bgug = null;
if (Train.cheatPCFG) {
List allTrees = new ArrayList(binaryTrainTrees);
allTrees.addAll(binaryTestTrees);
bgug = (Pair) bgExtractor.extract(allTrees);
} else {
bgug = (Pair) bgExtractor.extract(binaryTrainTrees);
}
bg = (BinaryGrammar) bgug.second;
bg.splitRules();
ug = (UnaryGrammar) bgug.first;
ug.purgeRules();
Timing.tick("done.");
}
System.err.print("Extracting Lexicon...");
lex = op.tlpParams.lex(op.lexOptions);
lex.train(binaryTrainTrees);
Timing.tick("done.");
if (op.doDep) {
System.err.print("Extracting Dependencies...");
binaryTrainTrees.clear();
// dgBLIPP = (DependencyGrammar) dgExtractor.extract(new
// ConcatenationIterator(trainTreebank.iterator(),blippTreebank.iterator()),new
// TransformTreeDependency(tlpParams,true));
DependencyGrammar dg1 =
(DependencyGrammar)
dgExtractor.extract(
trainTreebank.iterator(), new TransformTreeDependency(op.tlpParams, true));
// dgBLIPP=(DependencyGrammar)dgExtractor.extract(blippTreebank.iterator(),new
// TransformTreeDependency(tlpParams));
// dg = (DependencyGrammar) dgExtractor.extract(new
// ConcatenationIterator(trainTreebank.iterator(),blippTreebank.iterator()),new
// TransformTreeDependency(tlpParams));
// dg=new DependencyGrammarCombination(dg1,dgBLIPP,2);
// dg = (DependencyGrammar) dgExtractor.extract(binaryTrainTrees); //uses information whether
// the words are known or not, discards unknown words
Timing.tick("done.");
// System.out.print("Extracting Unknown Word Model...");
// UnknownWordModel uwm = (UnknownWordModel)uwmExtractor.extract(binaryTrainTrees);
// Timing.tick("done.");
System.out.print("Tuning Dependency Model...");
dg.tune(binaryTestTrees);
// System.out.println("TUNE DEPS: "+tuneDeps);
Timing.tick("done.");
}
BinaryGrammar boundBG = bg;
UnaryGrammar boundUG = ug;
GrammarProjection gp = new NullGrammarProjection(bg, ug);
// serialization
if (serializeFile != null) {
System.err.print("Serializing parser...");
LexicalizedParser.saveParserDataToSerialized(
new ParserData(lex, bg, ug, dg, Numberer.getNumberers(), op), serializeFile);
Timing.tick("done.");
}
// test: pcfg-parse and output
ExhaustivePCFGParser parser = null;
if (op.doPCFG) {
parser = new ExhaustivePCFGParser(boundBG, boundUG, lex, op);
}
ExhaustiveDependencyParser dparser =
((op.doDep && !Test.useFastFactored) ? new ExhaustiveDependencyParser(dg, lex, op) : null);
Scorer scorer = (op.doPCFG ? new TwinScorer(new ProjectionScorer(parser, gp), dparser) : null);
// Scorer scorer = parser;
BiLexPCFGParser bparser = null;
if (op.doPCFG && op.doDep) {
bparser =
(Test.useN5)
? new BiLexPCFGParser.N5BiLexPCFGParser(
scorer, parser, dparser, bg, ug, dg, lex, op, gp)
: new BiLexPCFGParser(scorer, parser, dparser, bg, ug, dg, lex, op, gp);
}
LabeledConstituentEval pcfgPE = new LabeledConstituentEval("pcfg PE", true, tlp);
LabeledConstituentEval comboPE = new LabeledConstituentEval("combo PE", true, tlp);
AbstractEval pcfgCB = new LabeledConstituentEval.CBEval("pcfg CB", true, tlp);
AbstractEval pcfgTE = new AbstractEval.TaggingEval("pcfg TE");
AbstractEval comboTE = new AbstractEval.TaggingEval("combo TE");
AbstractEval pcfgTEnoPunct = new AbstractEval.TaggingEval("pcfg nopunct TE");
AbstractEval comboTEnoPunct = new AbstractEval.TaggingEval("combo nopunct TE");
AbstractEval depTE = new AbstractEval.TaggingEval("depnd TE");
AbstractEval depDE =
new AbstractEval.DependencyEval("depnd DE", true, tlp.punctuationWordAcceptFilter());
AbstractEval comboDE =
new AbstractEval.DependencyEval("combo DE", true, tlp.punctuationWordAcceptFilter());
if (Test.evalb) {
EvalB.initEVALBfiles(op.tlpParams);
}
// int[] countByLength = new int[Test.maxLength+1];
// use a reflection ruse, so one can run this without needing the tagger
// edu.stanford.nlp.process.SentenceTagger tagger = (Test.preTag ? new
// edu.stanford.nlp.process.SentenceTagger("/u/nlp/data/tagger.params/wsj0-21.holder") : null);
SentenceProcessor tagger = null;
if (Test.preTag) {
try {
Class[] argsClass = new Class[] {String.class};
Object[] arguments =
new Object[] {"/u/nlp/data/pos-tagger/wsj3t0-18-bidirectional/train-wsj-0-18.holder"};
tagger =
(SentenceProcessor)
Class.forName("edu.stanford.nlp.tagger.maxent.MaxentTagger")
.getConstructor(argsClass)
.newInstance(arguments);
} catch (Exception e) {
System.err.println(e);
System.err.println("Warning: No pretagging of sentences will be done.");
}
}
for (int tNum = 0, ttSize = testTreebank.size(); tNum < ttSize; tNum++) {
Tree tree = testTreebank.get(tNum);
int testTreeLen = tree.yield().size();
if (testTreeLen > Test.maxLength) {
continue;
}
Tree binaryTree = binaryTestTrees.get(tNum);
// countByLength[testTreeLen]++;
System.out.println("-------------------------------------");
System.out.println("Number: " + (tNum + 1));
System.out.println("Length: " + testTreeLen);
// tree.pennPrint(pw);
// System.out.println("XXXX The binary tree is");
// binaryTree.pennPrint(pw);
// System.out.println("Here are the tags in the lexicon:");
// System.out.println(lex.showTags());
// System.out.println("Here's the tagnumberer:");
// System.out.println(Numberer.getGlobalNumberer("tags").toString());
long timeMil1 = System.currentTimeMillis();
Timing.tick("Starting parse.");
if (op.doPCFG) {
// System.err.println(Test.forceTags);
if (Test.forceTags) {
if (tagger != null) {
// System.out.println("Using a tagger to set tags");
// System.out.println("Tagged sentence as: " +
// tagger.processSentence(cutLast(wordify(binaryTree.yield()))).toString(false));
parser.parse(addLast(tagger.processSentence(cutLast(wordify(binaryTree.yield())))));
} else {
// System.out.println("Forcing tags to match input.");
parser.parse(cleanTags(binaryTree.taggedYield(), tlp));
}
} else {
// System.out.println("XXXX Parsing " + binaryTree.yield());
parser.parse(binaryTree.yield());
}
// Timing.tick("Done with pcfg phase.");
}
if (op.doDep) {
dparser.parse(binaryTree.yield());
// Timing.tick("Done with dependency phase.");
}
boolean bothPassed = false;
if (op.doPCFG && op.doDep) {
bothPassed = bparser.parse(binaryTree.yield());
// Timing.tick("Done with combination phase.");
}
long timeMil2 = System.currentTimeMillis();
long elapsed = timeMil2 - timeMil1;
System.err.println("Time: " + ((int) (elapsed / 100)) / 10.00 + " sec.");
// System.out.println("PCFG Best Parse:");
Tree tree2b = null;
Tree tree2 = null;
// System.out.println("Got full best parse...");
if (op.doPCFG) {
tree2b = parser.getBestParse();
tree2 = debinarizer.transformTree(tree2b);
}
// System.out.println("Debinarized parse...");
// tree2.pennPrint();
// System.out.println("DepG Best Parse:");
Tree tree3 = null;
Tree tree3db = null;
if (op.doDep) {
tree3 = dparser.getBestParse();
// was: but wrong Tree tree3db = debinarizer.transformTree(tree2);
tree3db = debinarizer.transformTree(tree3);
tree3.pennPrint(pw);
}
// tree.pennPrint();
// ((Tree)binaryTrainTrees.get(tNum)).pennPrint();
// System.out.println("Combo Best Parse:");
Tree tree4 = null;
if (op.doPCFG && op.doDep) {
try {
tree4 = bparser.getBestParse();
if (tree4 == null) {
tree4 = tree2b;
}
} catch (NullPointerException e) {
System.err.println("Blocked, using PCFG parse!");
tree4 = tree2b;
}
}
if (op.doPCFG && !bothPassed) {
tree4 = tree2b;
}
// tree4.pennPrint();
if (op.doDep) {
depDE.evaluate(tree3, binaryTree, pw);
depTE.evaluate(tree3db, tree, pw);
}
TreeTransformer tc = op.tlpParams.collinizer();
TreeTransformer tcEvalb = op.tlpParams.collinizerEvalb();
Tree tree4b = null;
if (op.doPCFG) {
// System.out.println("XXXX Best PCFG was: ");
// tree2.pennPrint();
// System.out.println("XXXX Transformed best PCFG is: ");
// tc.transformTree(tree2).pennPrint();
// System.out.println("True Best Parse:");
// tree.pennPrint();
// tc.transformTree(tree).pennPrint();
pcfgPE.evaluate(tc.transformTree(tree2), tc.transformTree(tree), pw);
pcfgCB.evaluate(tc.transformTree(tree2), tc.transformTree(tree), pw);
if (op.doDep) {
comboDE.evaluate((bothPassed ? tree4 : tree3), binaryTree, pw);
tree4b = tree4;
tree4 = debinarizer.transformTree(tree4);
if (op.nodePrune) {
NodePruner np = new NodePruner(parser, debinarizer);
tree4 = np.prune(tree4);
}
// tree4.pennPrint();
comboPE.evaluate(tc.transformTree(tree4), tc.transformTree(tree), pw);
}
// pcfgTE.evaluate(tree2, tree);
pcfgTE.evaluate(tcEvalb.transformTree(tree2), tcEvalb.transformTree(tree), pw);
pcfgTEnoPunct.evaluate(tc.transformTree(tree2), tc.transformTree(tree), pw);
if (op.doDep) {
comboTE.evaluate(tcEvalb.transformTree(tree4), tcEvalb.transformTree(tree), pw);
comboTEnoPunct.evaluate(tc.transformTree(tree4), tc.transformTree(tree), pw);
}
System.out.println("PCFG only: " + parser.scoreBinarizedTree(tree2b, 0));
// tc.transformTree(tree2).pennPrint();
tree2.pennPrint(pw);
if (op.doDep) {
System.out.println("Combo: " + parser.scoreBinarizedTree(tree4b, 0));
// tc.transformTree(tree4).pennPrint(pw);
tree4.pennPrint(pw);
}
System.out.println("Correct:" + parser.scoreBinarizedTree(binaryTree, 0));
/*
if (parser.scoreBinarizedTree(tree2b,true) < parser.scoreBinarizedTree(binaryTree,true)) {
System.out.println("SCORE INVERSION");
parser.validateBinarizedTree(binaryTree,0);
}
*/
tree.pennPrint(pw);
} // end if doPCFG
if (Test.evalb) {
if (op.doPCFG && op.doDep) {
EvalB.writeEVALBline(tcEvalb.transformTree(tree), tcEvalb.transformTree(tree4));
} else if (op.doPCFG) {
EvalB.writeEVALBline(tcEvalb.transformTree(tree), tcEvalb.transformTree(tree2));
} else if (op.doDep) {
EvalB.writeEVALBline(tcEvalb.transformTree(tree), tcEvalb.transformTree(tree3db));
}
}
} // end for each tree in test treebank
if (Test.evalb) {
EvalB.closeEVALBfiles();
}
// Test.display();
if (op.doPCFG) {
pcfgPE.display(false, pw);
System.out.println("Grammar size: " + Numberer.getGlobalNumberer("states").total());
pcfgCB.display(false, pw);
if (op.doDep) {
comboPE.display(false, pw);
}
pcfgTE.display(false, pw);
pcfgTEnoPunct.display(false, pw);
if (op.doDep) {
comboTE.display(false, pw);
comboTEnoPunct.display(false, pw);
}
}
if (op.doDep) {
depTE.display(false, pw);
depDE.display(false, pw);
}
if (op.doPCFG && op.doDep) {
comboDE.display(false, pw);
}
// pcfgPE.printGoodBad();
}
