/**
* Trains this lexicon on the Collection of trees. Also trains the unknown word model pointed to
* by this lexicon.
*/
public void train(Collection<Tree> trees, double weight, boolean keepTagsAsLabels) {
getUnknownWordModel().train(trees);
// scan data
for (Tree tree : trees) {
List<IntTaggedWord> taggedWords = treeToEvents(tree, keepTagsAsLabels);
for (int w = 0, sz = taggedWords.size(); w < sz; w++) {
IntTaggedWord iTW = taggedWords.get(w);
seenCounter.incrementCount(iTW, weight);
IntTaggedWord iT = new IntTaggedWord(nullWord, iTW.tag);
seenCounter.incrementCount(iT, weight);
IntTaggedWord iW = new IntTaggedWord(iTW.word, nullTag);
seenCounter.incrementCount(iW, weight);
IntTaggedWord i = new IntTaggedWord(nullWord, nullTag);
seenCounter.incrementCount(i, weight);
// rules.add(iTW);
tags.add(iT);
words.add(iW);
}
}
tune(trees);
// index the possible tags for each word
initRulesWithWord();
if (DEBUG_LEXICON) {
printLexStats();
}
}
@Override
public void evaluate(Tree t1, Tree t2, PrintWriter pw) {
Set<String> s1 = makeObjects(t1);
Set<String> s2 = makeObjects(t2);
for (String o1 : s1) {
if (!s2.contains(o1)) {
over.incrementCount(o1);
}
}
for (String o2 : s2) {
if (!s1.contains(o2)) {
under.incrementCount(o2);
}
}
}
public static <T> Counter<T> toCounter(List<FeatureValue<T>> featureValues) {
ClassicCounter<T> counter = new ClassicCounter<T>();
for (FeatureValue<T> fv : featureValues) {
counter.incrementCount(fv.name, fv.value);
}
return counter;
}
@Override
protected void tallyRoot(Tree lt, double weight) {
// this list is in full (not reduced) tag space
List<IntDependency> deps = MLEDependencyGrammar.treeToDependencyList(lt, wordIndex, tagIndex);
for (IntDependency dependency : deps) {
dependencyCounter.incrementCount(dependency, weight);
}
}
@Override
public void evaluate(Tree t1, Tree t2, PrintWriter pw) {
List<String> s1 = myMakeObjects(t1);
List<String> s2 = myMakeObjects(t2);
List<String> del2 = new LinkedList<>(s2);
// we delete out as we find them so we can score correctly a cat with
// a certain cardinality in a tree.
for (String o1 : s1) {
if (!del2.remove(o1)) {
over.incrementCount(o1);
}
}
for (String o2 : s2) {
if (!s1.remove(o2)) {
under.incrementCount(o2);
}
}
}
/** Make a copy of the array of counters. */
public ClassicCounter<Integer>[] cloneCounter(ClassicCounter<Integer>[] counter) {
ClassicCounter<Integer>[] newcount =
ErasureUtils.<ClassicCounter<Integer>>mkTArray(ClassicCounter.class, counter.length);
for (int xx = 0; xx < counter.length; xx++) {
ClassicCounter<Integer> cc = new ClassicCounter<Integer>();
newcount[xx] = cc;
for (Integer key : counter[xx].keySet()) cc.incrementCount(key, counter[xx].getCount(key));
}
return newcount;
}
private void expandStop(
IntDependency dependency, short distBinDist, double count, boolean wildForStop) {
IntTaggedWord headT = getCachedITW(dependency.head.tag);
IntTaggedWord head =
new IntTaggedWord(dependency.head.word, tagBin(dependency.head.tag)); // dependency.head;
IntTaggedWord arg =
new IntTaggedWord(dependency.arg.word, tagBin(dependency.arg.tag)); // dependency.arg;
boolean leftHeaded = dependency.leftHeaded;
if (arg.word == STOP_WORD_INT) {
stopCounter.incrementCount(intern(head, arg, leftHeaded, distBinDist), count);
stopCounter.incrementCount(intern(headT, arg, leftHeaded, distBinDist), count);
}
if (wildForStop || arg.word != STOP_WORD_INT) {
stopCounter.incrementCount(intern(head, wildTW, leftHeaded, distBinDist), count);
stopCounter.incrementCount(intern(headT, wildTW, leftHeaded, distBinDist), count);
}
}
/** @param <T> */
public static <T> List<FeatureValue<T>> combine(Collection<FeatureValue<T>> featureValues) {
ClassicCounter<T> counter = new ClassicCounter<T>();
for (FeatureValue<T> fv : featureValues) {
counter.incrementCount(fv.name, fv.value);
}
Set<T> keys = new TreeSet<T>(counter.keySet());
List<FeatureValue<T>> featureList = new ArrayList<FeatureValue<T>>(keys.size());
for (T key : keys) {
featureList.add(new FeatureValue<T>(key, counter.getCount(key)));
}
return featureList;
}
public UnknownWordModel finishTraining() {
// make sure the unseen counter isn't empty! If it is, put in
// a uniform unseen over tags
if (unSeenCounter.isEmpty()) {
int numTags = tagIndex.size();
for (int tt = 0; tt < numTags; tt++) {
if (!Lexicon.BOUNDARY_TAG.equals(tagIndex.get(tt))) {
IntTaggedWord iT = new IntTaggedWord(nullWord, tt);
IntTaggedWord i = NULL_ITW;
unSeenCounter.incrementCount(iT);
unSeenCounter.incrementCount(i);
}
}
}
// index the possible tags for each word
// numWords = wordIndex.size();
// unknownWordIndex = wordIndex.indexOf(Lexicon.UNKNOWN_WORD, true);
// initRulesWithWord();
return model;
}
private void writeObject(ObjectOutputStream stream) throws IOException {
// System.err.println("\nBefore compression:");
// System.err.println("arg size: " + argCounter.size() + " total: " +
// argCounter.totalCount());
// System.err.println("stop size: " + stopCounter.size() + " total: " +
// stopCounter.totalCount());
ClassicCounter<IntDependency> fullArgCounter = argCounter;
argCounter = new ClassicCounter<IntDependency>();
for (IntDependency dependency : fullArgCounter.keySet()) {
if (dependency.head != wildTW
&& dependency.arg != wildTW
&& dependency.head.word != -1
&& dependency.arg.word != -1) {
argCounter.incrementCount(dependency, fullArgCounter.getCount(dependency));
}
}
ClassicCounter<IntDependency> fullStopCounter = stopCounter;
stopCounter = new ClassicCounter<IntDependency>();
for (IntDependency dependency : fullStopCounter.keySet()) {
if (dependency.head.word != -1) {
stopCounter.incrementCount(dependency, fullStopCounter.getCount(dependency));
}
}
// System.err.println("After compression:");
// System.err.println("arg size: " + argCounter.size() + " total: " +
// argCounter.totalCount());
// System.err.println("stop size: " + stopCounter.size() + " total: " +
// stopCounter.totalCount());
stream.defaultWriteObject();
argCounter = fullArgCounter;
stopCounter = fullStopCounter;
}
/** Trains this UWM on the Collection of trees. */
public void train(TaggedWord tw, int loc, double weight) {
IntTaggedWord iTW = new IntTaggedWord(tw.word(), tw.tag(), wordIndex, tagIndex);
IntTaggedWord iT = new IntTaggedWord(nullWord, iTW.tag);
IntTaggedWord iW = new IntTaggedWord(iTW.word, nullTag);
seenCounter.incrementCount(iW, weight);
IntTaggedWord i = NULL_ITW;
if (treesRead > indexToStartUnkCounting) {
// start doing this once some way through trees;
// treesRead is 1 based counting
if (seenCounter.getCount(iW) < 1.5) {
// it's an entirely unknown word
int s = model.getSignatureIndex(iTW.word, loc, wordIndex.get(iTW.word));
if (DOCUMENT_UNKNOWNS) {
String wStr = wordIndex.get(iTW.word);
String tStr = tagIndex.get(iTW.tag);
String sStr = wordIndex.get(s);
EncodingPrintWriter.err.println(
"Unknown word/tag/sig:\t" + wStr + '\t' + tStr + '\t' + sStr, "UTF-8");
}
IntTaggedWord iTS = new IntTaggedWord(s, iTW.tag);
IntTaggedWord iS = new IntTaggedWord(s, nullTag);
unSeenCounter.incrementCount(iTS, weight);
unSeenCounter.incrementCount(iT, weight);
unSeenCounter.incrementCount(iS, weight);
unSeenCounter.incrementCount(i, weight);
// rules.add(iTS);
// sigs.add(iS);
} // else {
// if (seenCounter.getCount(iTW) < 2) {
// it's a new tag for a known word
// do nothing for now
// }
// }
}
}
@Override
public void train(List<TaggedWord> sentence) {
lex.train(sentence, 1.0);
String last = null;
for (TaggedWord tagLabel : sentence) {
String tag = tagLabel.tag();
tagIndex.add(tag);
if (last == null) {
initial.incrementCount(tag);
} else {
ruleCounter.incrementCount2D(last, tag);
}
last = tag;
}
}
/** Adds the tagging with count to the data structures in this Lexicon. */
protected void addTagging(boolean seen, IntTaggedWord itw, double count) {
if (seen) {
seenCounter.incrementCount(itw, count);
if (itw.tag() == nullTag) {
words.add(itw);
} else if (itw.word() == nullWord) {
tags.add(itw);
} else {
// rules.add(itw);
}
} else {
uwModel.addTagging(seen, itw, count);
// if (itw.tag() == nullTag) {
// sigs.add(itw);
// }
}
}
private Distribution<Integer> getSegmentedWordLengthDistribution(Treebank tb) {
// CharacterLevelTagExtender ext = new CharacterLevelTagExtender();
ClassicCounter<Integer> c = new ClassicCounter<Integer>();
for (Iterator iterator = tb.iterator(); iterator.hasNext(); ) {
Tree gold = (Tree) iterator.next();
StringBuilder goldChars = new StringBuilder();
ArrayList goldYield = gold.yield();
for (Iterator wordIter = goldYield.iterator(); wordIter.hasNext(); ) {
Word word = (Word) wordIter.next();
goldChars.append(word);
}
List<HasWord> ourWords = segment(goldChars.toString());
for (int i = 0; i < ourWords.size(); i++) {
c.incrementCount(Integer.valueOf(ourWords.get(i).word().length()));
}
}
return Distribution.getDistribution(c);
}
public ClassicCounter<L> scoresOf(RVFDatum<L, F> example) {
ClassicCounter<L> scores = new ClassicCounter<>();
Counters.addInPlace(scores, priors);
if (addZeroValued) {
Counters.addInPlace(scores, priorZero);
}
for (L l : labels) {
double score = 0.0;
Counter<F> features = example.asFeaturesCounter();
for (F f : features.keySet()) {
int value = (int) features.getCount(f);
score += weight(l, f, Integer.valueOf(value));
if (addZeroValued) {
score -= weight(l, f, zero);
}
}
scores.incrementCount(l, score);
}
return scores;
}
/**
* Collect counts for a non-STOP dependent. The dependency arg is still in the full tag space.
*
* @param dependency A non-stop dependency
* @param valBinDist A binned distance
* @param count The weight with which to add this dependency
*/
private void expandArg(IntDependency dependency, short valBinDist, double count) {
IntTaggedWord headT = getCachedITW(dependency.head.tag);
IntTaggedWord argT = getCachedITW(dependency.arg.tag);
IntTaggedWord head =
new IntTaggedWord(dependency.head.word, tagBin(dependency.head.tag)); // dependency.head;
IntTaggedWord arg =
new IntTaggedWord(dependency.arg.word, tagBin(dependency.arg.tag)); // dependency.arg;
boolean leftHeaded = dependency.leftHeaded;
// argCounter stores stuff in both the original and the reduced tag space???
argCounter.incrementCount(intern(head, arg, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headT, arg, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(head, argT, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headT, argT, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(head, wildTW, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headT, wildTW, leftHeaded, valBinDist), count);
// the WILD head stats are always directionless and not useDistance!
argCounter.incrementCount(intern(wildTW, arg, false, (short) -1), count);
argCounter.incrementCount(intern(wildTW, argT, false, (short) -1), count);
if (useSmoothTagProjection) {
// added stuff to do more smoothing. CDM Jan 2007
IntTaggedWord headP =
new IntTaggedWord(dependency.head.word, tagProject(dependency.head.tag));
IntTaggedWord headTP = new IntTaggedWord(ANY_WORD_INT, tagProject(dependency.head.tag));
IntTaggedWord argP = new IntTaggedWord(dependency.arg.word, tagProject(dependency.arg.tag));
IntTaggedWord argTP = new IntTaggedWord(ANY_WORD_INT, tagProject(dependency.arg.tag));
argCounter.incrementCount(intern(headP, argP, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headTP, argP, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headP, argTP, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headTP, argTP, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headP, wildTW, leftHeaded, valBinDist), count);
argCounter.incrementCount(intern(headTP, wildTW, leftHeaded, valBinDist), count);
// the WILD head stats are always directionless and not useDistance!
argCounter.incrementCount(intern(wildTW, argP, false, (short) -1), count);
argCounter.incrementCount(intern(wildTW, argTP, false, (short) -1), count);
argCounter.incrementCount(
intern(wildTW, new IntTaggedWord(dependency.head.word, ANY_TAG_INT), false, (short) -1),
count);
}
numWordTokens++;
}
