private void incrementMonth(ISODateInstance referenceDate, Pair<DateField, Integer> relation) {
String origDateString = referenceDate.getStartDate();
String monthString = origDateString.substring(4, 6);
if (monthString.contains("*")) {
isoDate = origDateString;
return;
}
// Month is not a variable
Integer monthNum = Integer.parseInt(monthString);
// Check if we're an edge case
if (((monthNum + relation.second()) > 12) || ((monthNum + relation.second) < 1)) {
boolean decreasing = ((monthNum + relation.second) < 1);
int newMonthNum = (monthNum + relation.second()) % 12;
if (newMonthNum < 0) {
newMonthNum *= -1;
}
// Set the month appropriately
isoDate = makeStringMonthChange(origDateString, newMonthNum);
// Increment the year if possible
String yearString = origDateString.substring(0, 4);
if (!yearString.contains("*")) {
// How much we increment depends on above mod
int numYearsToIncrement = (int) Math.ceil(relation.second() / 12.0);
if (decreasing) {
isoDate =
makeStringYearChange(isoDate, Integer.parseInt(yearString) - numYearsToIncrement);
} else {
isoDate =
makeStringYearChange(isoDate, Integer.parseInt(yearString) + numYearsToIncrement);
}
}
} else {
isoDate = makeStringMonthChange(origDateString, (monthNum + relation.second()));
}
}
public static Tree processPatternsOnTree(List<Pair<TregexPattern, TsurgeonPattern>> ops, Tree t) {
matchedOnTree = false;
for (Pair<TregexPattern, TsurgeonPattern> op : ops) {
try {
if (DEBUG) {
System.err.println("Running pattern " + op.first());
}
TregexMatcher m = op.first().matcher(t);
while (m.find()) {
matchedOnTree = true;
t = op.second().evaluate(t, m);
if (t == null) {
return null;
}
m = op.first().matcher(t);
}
} catch (NullPointerException npe) {
throw new RuntimeException(
"Tsurgeon.processPatternsOnTree failed to match label for pattern: "
+ op.first()
+ ", "
+ op.second(),
npe);
}
}
return t;
}
/**
* Uses regexp matching to match month, day, and year fields TODO: Find a way to mark what;s
* already been handled in the string
*/
public boolean extractFields(String inputDate) {
if (tokens.size() < 2) {
tokenizeDate(inputDate);
}
if (DEBUG) {
System.err.println("Extracting date: " + inputDate);
}
// first we see if it's a hyphen and two parseable dates - if not, we treat it as one date
Pair<String, String> dateEndpoints = getRangeDates(inputDate);
if (dateEndpoints != null) {
ISODateInstance date1 = new ISODateInstance(dateEndpoints.first());
if (dateEndpoints.first().contains(" ") && !dateEndpoints.second().contains(" ")) {
// consider whether it's a leading modifier; e.g., "June 8-10" will be split into June 8,
// and 10 when really we'd like June 8 and June 10
String date =
dateEndpoints.first().substring(0, dateEndpoints.first().indexOf(' '))
+ ' '
+ dateEndpoints.second();
ISODateInstance date2 = new ISODateInstance(date);
if (!date1.isUnparseable() && !date2.isUnparseable()) {
isoDate = (new ISODateInstance(date1, date2)).getDateString();
return true;
}
}
ISODateInstance date2 = new ISODateInstance(dateEndpoints.second());
if (!date1.isUnparseable() && !date2.isUnparseable()) {
isoDate = (new ISODateInstance(date1, date2)).getDateString();
return true;
}
}
if (extractYYYYMMDD(inputDate)) {
return true;
}
if (extractMMDDYY(inputDate)) {
return true;
}
boolean passed = false;
passed = extractYear(inputDate) || passed;
passed = extractMonth(inputDate) || passed;
passed = extractDay(inputDate) || passed;
// slightly hacky, but check for some common modifiers that get grouped into the date
passed = addExtraRanges(inputDate) || passed;
if (!passed) { // couldn't parse
// try one more trick
unparseable = true;
boolean weekday = extractWeekday(inputDate);
if (!weekday) {
isoDate = inputDate;
}
}
return passed;
}
private void commitVariableGroups(Matcher m) {
committedVariables = true; // commit all my variable groups.
for (Pair<Integer, String> varGroup : myNode.variableGroups) {
String thisVarString = m.group(varGroup.first());
variableStrings.setVar(varGroup.second(), thisVarString);
}
}
/** Check one mention is the speaker of the other mention */
public static boolean isSpeaker(Mention m, Mention ant, Dictionaries dict) {
if (!dict.firstPersonPronouns.contains(ant.spanToString().toLowerCase())
|| ant.number == Number.PLURAL
|| ant.sentNum != m.sentNum) return false;
int countQuotationMark = 0;
for (int i = Math.min(m.headIndex, ant.headIndex) + 1;
i < Math.max(m.headIndex, ant.headIndex);
i++) {
String word = m.sentenceWords.get(i).get(CoreAnnotations.TextAnnotation.class);
if (word.equals("``") || word.equals("''")) countQuotationMark++;
}
if (countQuotationMark != 1) return false;
IndexedWord w =
m.dependency.getNodeByWordPattern(
m.sentenceWords.get(m.headIndex).get(CoreAnnotations.TextAnnotation.class));
if (w == null) return false;
for (Pair<GrammaticalRelation, IndexedWord> parent : m.dependency.parentPairs(w)) {
if (parent.first().getShortName().equals("nsubj")
&& dict.reportVerb.contains(parent.second().get(CoreAnnotations.LemmaAnnotation.class))) {
return true;
}
}
return false;
}
private String findNextParagraphSpeaker(
List<CoreMap> paragraph, int paragraphOffset, Dictionaries dict) {
CoreMap lastSent = paragraph.get(paragraph.size() - 1);
String speaker = "";
for (CoreLabel w : lastSent.get(CoreAnnotations.TokensAnnotation.class)) {
if (w.get(CoreAnnotations.LemmaAnnotation.class).equals("report")
|| w.get(CoreAnnotations.LemmaAnnotation.class).equals("say")) {
String word = w.get(CoreAnnotations.TextAnnotation.class);
SemanticGraph dependency =
lastSent.get(SemanticGraphCoreAnnotations.CollapsedDependenciesAnnotation.class);
IndexedWord t = dependency.getNodeByWordPattern(word);
for (Pair<GrammaticalRelation, IndexedWord> child : dependency.childPairs(t)) {
if (child.first().getShortName().equals("nsubj")) {
int subjectIndex = child.second().index(); // start from 1
IntTuple headPosition = new IntTuple(2);
headPosition.set(0, paragraph.size() - 1 + paragraphOffset);
headPosition.set(1, subjectIndex - 1);
if (mentionheadPositions.containsKey(headPosition)
&& mentionheadPositions.get(headPosition).nerString.startsWith("PER")) {
speaker = Integer.toString(mentionheadPositions.get(headPosition).mentionID);
}
}
}
}
}
return speaker;
}
private void decommitVariableGroups() {
if (committedVariables) {
for (Pair<Integer, String> varGroup : myNode.variableGroups) {
variableStrings.unsetVar(varGroup.second());
}
}
committedVariables = false;
}
private boolean findSpeaker(
int utterNum,
int sentNum,
List<CoreMap> sentences,
int startIndex,
int endIndex,
Dictionaries dict) {
List<CoreLabel> sent = sentences.get(sentNum).get(CoreAnnotations.TokensAnnotation.class);
for (int i = startIndex; i < endIndex; i++) {
if (sent.get(i).get(CoreAnnotations.UtteranceAnnotation.class) != 0) continue;
String lemma = sent.get(i).get(CoreAnnotations.LemmaAnnotation.class);
String word = sent.get(i).get(CoreAnnotations.TextAnnotation.class);
if (dict.reportVerb.contains(lemma)) {
// find subject
SemanticGraph dependency =
sentences
.get(sentNum)
.get(SemanticGraphCoreAnnotations.CollapsedDependenciesAnnotation.class);
IndexedWord w = dependency.getNodeByWordPattern(word);
if (w != null) {
for (Pair<GrammaticalRelation, IndexedWord> child : dependency.childPairs(w)) {
if (child.first().getShortName().equals("nsubj")) {
String subjectString = child.second().word();
int subjectIndex = child.second().index(); // start from 1
IntTuple headPosition = new IntTuple(2);
headPosition.set(0, sentNum);
headPosition.set(1, subjectIndex - 1);
String speaker;
if (mentionheadPositions.containsKey(headPosition)) {
speaker = Integer.toString(mentionheadPositions.get(headPosition).mentionID);
} else {
speaker = subjectString;
}
speakers.put(utterNum, speaker);
return true;
}
}
} else {
SieveCoreferenceSystem.logger.warning("Cannot find node in dependency for word " + word);
}
}
}
return false;
}
private void findQuotationSpeaker(
int utterNum,
List<CoreMap> sentences,
Pair<Integer, Integer> beginQuotation,
Pair<Integer, Integer> endQuotation,
Dictionaries dict) {
if (findSpeaker(utterNum, beginQuotation.first(), sentences, 0, beginQuotation.second(), dict))
return;
if (findSpeaker(
utterNum,
endQuotation.first(),
sentences,
endQuotation.second(),
sentences.get(endQuotation.first()).get(CoreAnnotations.TokensAnnotation.class).size(),
dict)) return;
if (beginQuotation.second() <= 1 && beginQuotation.first() > 0) {
if (findSpeaker(
utterNum,
beginQuotation.first() - 1,
sentences,
0,
sentences
.get(beginQuotation.first() - 1)
.get(CoreAnnotations.TokensAnnotation.class)
.size(),
dict)) return;
}
if (endQuotation.second() == sentences.get(endQuotation.first()).size() - 1
&& sentences.size() > endQuotation.first() + 1) {
if (findSpeaker(
utterNum,
endQuotation.first() + 1,
sentences,
0,
sentences
.get(endQuotation.first() + 1)
.get(CoreAnnotations.TokensAnnotation.class)
.size(),
dict)) return;
}
}
/**
* Compares this <code>Pair</code> to another object. If the object is a <code>Pair</code>, this
* function will work providing the elements of the <code>Pair</code> are themselves comparable.
* It will then return a value based on the pair of objects, where <code>
* p > q iff p.first() > q.first() ||
* (p.first().equals(q.first()) && p.second() > q.second())</code>. If the other object is not
* a <code>Pair</code>, it throws a <code>ClassCastException</code>.
*
* @param o the <code>Object</code> to be compared.
* @return the value <code>0</code> if the argument is a <code>Pair</code> equal to this <code>
* Pair</code>; a value less than <code>0</code> if the argument is a <code>Pair</code>
* greater than this <code>Pair</code>; and a value greater than <code>0</code> if the
* argument is a <code>Pair</code> less than this <code>Pair</code>.
* @throws ClassCastException if the argument is not a <code>Pair</code>.
* @see java.lang.Comparable
*/
public int compareTo(Object o) {
Pair another = (Pair) o;
int comp = ((Comparable) first()).compareTo(another.first());
if (comp != 0) {
return comp;
} else {
return ((Comparable) second()).compareTo(another.second());
}
}
public String expandStringRegex(String regex) {
// Replace all variables in regex
String expanded = regex;
for (String v : stringRegexVariables.keySet()) {
Pair<Pattern, String> p = stringRegexVariables.get(v);
expanded = p.first().matcher(expanded).replaceAll(p.second());
}
return expanded;
}
/**
* Read a string representation of a Pair from a DataStream. This might not work correctly unless
* the pair of objects are of type <code>String</code>.
*/
public static Pair<String, String> readStringPair(DataInputStream in) {
Pair<String, String> p = new Pair<String, String>();
try {
p.first = in.readUTF();
p.second = in.readUTF();
} catch (Exception e) {
e.printStackTrace();
}
return p;
}
private void incrementYear(ISODateInstance referenceDate, Pair<DateField, Integer> relation) {
String origDateString = referenceDate.getStartDate();
String yearString = origDateString.substring(0, 4);
if (yearString.contains("*")) {
isoDate = origDateString;
return;
}
isoDate =
makeStringYearChange(origDateString, Integer.parseInt(yearString) + relation.second());
}
private static void extractSubtrees(List<String> codeStrings, String treeFile) {
List<Pair<Integer, Integer>> codes = new ArrayList<Pair<Integer, Integer>>();
for (String s : codeStrings) {
Matcher m = codePattern.matcher(s);
if (m.matches())
codes.add(
new Pair<Integer, Integer>(Integer.parseInt(m.group(1)), Integer.parseInt(m.group(2))));
else throw new RuntimeException("Error: illegal node code " + s);
}
TreeReaderFactory trf = new TRegexTreeReaderFactory();
MemoryTreebank treebank = new MemoryTreebank(trf);
treebank.loadPath(treeFile, null, true);
for (Pair<Integer, Integer> code : codes) {
Tree t = treebank.get(code.first() - 1);
t.getNodeNumber(code.second()).pennPrint();
}
}
/**
* Returns true if there is a feasible combination of child branch ids that causes all child
* expressions to be satisfied with respect to the specified child expression (assuming
* satisfiction with the specified branch and node index) For other child expressions to have a
* compatible satisfiable branch, that branch must also terminate with the same node index as
* this one.
*
* @param index - Index of the child expression
* @param bid - Branch id that causes the indexed child to be satisfied
* @param pos - Node index that causes the indexed child to be satisfied
* @return whether there is a feasible combination that causes all children to be satisfied with
* respect to specfied child.
*/
private boolean isAllChildMatched(int index, int bid, int pos) {
for (int i = 0; i < reachableChildBids.length; i++) {
Set<Pair<Integer, Integer>> v = reachableChildBids[i];
if (v == null || v.isEmpty()) return false;
if (i != index) {
boolean ok = false;
for (Pair<Integer, Integer> p : v) {
if (p.second() == pos) {
ok = true;
break;
}
}
if (!ok) {
return false;
}
}
}
return true;
}
/**
* Returns array of child branch ids that causes all child expressions to be satisfied with
* respect to the specified child expression (assuming satisfiction with the specified branch
* and node index) For other child expressions to have a compatible satisfiable branch, that
* branch must also terminate with the same node index as this one.
*
* @param index - Index of the child expression
* @param bid - Branch id that causes the indexed child to be satisfied
* @param pos - Node index that causes the indexed child to be satisfied
* @return array of child branch ids if there is a valid combination null otherwise
*/
private int[] getAllChildMatchedBids(int index, int bid, int pos) {
int[] matchedBids = new int[reachableChildBids.length];
for (int i = 0; i < reachableChildBids.length; i++) {
Set<Pair<Integer, Integer>> v = reachableChildBids[i];
if (v == null || v.isEmpty()) return null;
if (i != index) {
boolean ok = false;
for (Pair<Integer, Integer> p : v) {
if (p.second() == pos) {
ok = true;
matchedBids[i] = p.first();
break;
}
}
if (!ok) {
return null;
}
} else {
matchedBids[i] = bid;
}
}
return matchedBids;
}
protected <T> boolean match(
int bid, SequenceMatcher.MatchedStates<T> matchedStates, boolean consume) {
// Try to match previous node/nodes exactly
if (consume) {
// First element is group that is matched, second is number of nodes matched so far
Pair<SequenceMatcher.MatchedGroup, Integer> backRefState =
(Pair<SequenceMatcher.MatchedGroup, Integer>)
matchedStates.getBranchStates().getMatchStateInfo(bid, this);
if (backRefState == null) {
// Haven't tried to match this node before, try now
// Get element and return if it matched or not
SequenceMatcher.MatchedGroup matchedGroup =
matchedStates.getBranchStates().getMatchedGroup(bid, captureGroupId);
if (matchedGroup != null) {
// See if the first node matches
if (matchedGroup.matchEnd > matchedGroup.matchBegin) {
boolean matched = match(bid, matchedStates, matchedGroup, 0);
return matched;
} else {
// TODO: Check handling of previous nodes that are zero elements?
return super.match(bid, matchedStates, consume);
}
}
return false;
} else {
SequenceMatcher.MatchedGroup matchedGroup = backRefState.first();
int matchedNodes = backRefState.second();
boolean matched = match(bid, matchedStates, matchedGroup, matchedNodes);
return matched;
}
} else {
// Not consuming, just add this state back to list of states to be processed
matchedStates.addState(bid, this);
return false;
}
}
/** @param args */
public static void main(String[] args) {
if (args.length != 3) {
System.err.printf(
"Usage: java %s language filename features%n",
TreebankFactoredLexiconStats.class.getName());
System.exit(-1);
}
Language language = Language.valueOf(args[0]);
TreebankLangParserParams tlpp = language.params;
if (language.equals(Language.Arabic)) {
String[] options = {"-arabicFactored"};
tlpp.setOptionFlag(options, 0);
} else {
String[] options = {"-frenchFactored"};
tlpp.setOptionFlag(options, 0);
}
Treebank tb = tlpp.diskTreebank();
tb.loadPath(args[1]);
MorphoFeatureSpecification morphoSpec =
language.equals(Language.Arabic)
? new ArabicMorphoFeatureSpecification()
: new FrenchMorphoFeatureSpecification();
String[] features = args[2].trim().split(",");
for (String feature : features) {
morphoSpec.activate(MorphoFeatureType.valueOf(feature));
}
// Counters
Counter<String> wordTagCounter = new ClassicCounter<>(30000);
Counter<String> morphTagCounter = new ClassicCounter<>(500);
// Counter<String> signatureTagCounter = new ClassicCounter<String>();
Counter<String> morphCounter = new ClassicCounter<>(500);
Counter<String> wordCounter = new ClassicCounter<>(30000);
Counter<String> tagCounter = new ClassicCounter<>(300);
Counter<String> lemmaCounter = new ClassicCounter<>(25000);
Counter<String> lemmaTagCounter = new ClassicCounter<>(25000);
Counter<String> richTagCounter = new ClassicCounter<>(1000);
Counter<String> reducedTagCounter = new ClassicCounter<>(500);
Counter<String> reducedTagLemmaCounter = new ClassicCounter<>(500);
Map<String, Set<String>> wordLemmaMap = Generics.newHashMap();
TwoDimensionalIntCounter<String, String> lemmaReducedTagCounter =
new TwoDimensionalIntCounter<>(30000);
TwoDimensionalIntCounter<String, String> reducedTagTagCounter =
new TwoDimensionalIntCounter<>(500);
TwoDimensionalIntCounter<String, String> tagReducedTagCounter =
new TwoDimensionalIntCounter<>(300);
int numTrees = 0;
for (Tree tree : tb) {
for (Tree subTree : tree) {
if (!subTree.isLeaf()) {
tlpp.transformTree(subTree, tree);
}
}
List<Label> pretermList = tree.preTerminalYield();
List<Label> yield = tree.yield();
assert yield.size() == pretermList.size();
int yieldLen = yield.size();
for (int i = 0; i < yieldLen; ++i) {
String tag = pretermList.get(i).value();
String word = yield.get(i).value();
String morph = ((CoreLabel) yield.get(i)).originalText();
// Note: if there is no lemma, then we use the surface form.
Pair<String, String> lemmaTag = MorphoFeatureSpecification.splitMorphString(word, morph);
String lemma = lemmaTag.first();
String richTag = lemmaTag.second();
// WSGDEBUG
if (tag.contains("MW")) lemma += "-MWE";
lemmaCounter.incrementCount(lemma);
lemmaTagCounter.incrementCount(lemma + tag);
richTagCounter.incrementCount(richTag);
String reducedTag = morphoSpec.strToFeatures(richTag).toString();
reducedTagCounter.incrementCount(reducedTag);
reducedTagLemmaCounter.incrementCount(reducedTag + lemma);
wordTagCounter.incrementCount(word + tag);
morphTagCounter.incrementCount(morph + tag);
morphCounter.incrementCount(morph);
wordCounter.incrementCount(word);
tagCounter.incrementCount(tag);
reducedTag = reducedTag.equals("") ? "NONE" : reducedTag;
if (wordLemmaMap.containsKey(word)) {
wordLemmaMap.get(word).add(lemma);
} else {
Set<String> lemmas = Generics.newHashSet(1);
wordLemmaMap.put(word, lemmas);
}
lemmaReducedTagCounter.incrementCount(lemma, reducedTag);
reducedTagTagCounter.incrementCount(lemma + reducedTag, tag);
tagReducedTagCounter.incrementCount(tag, reducedTag);
}
++numTrees;
}
// Barf...
System.out.println("Language: " + language.toString());
System.out.printf("#trees:\t%d%n", numTrees);
System.out.printf("#tokens:\t%d%n", (int) wordCounter.totalCount());
System.out.printf("#words:\t%d%n", wordCounter.keySet().size());
System.out.printf("#tags:\t%d%n", tagCounter.keySet().size());
System.out.printf("#wordTagPairs:\t%d%n", wordTagCounter.keySet().size());
System.out.printf("#lemmas:\t%d%n", lemmaCounter.keySet().size());
System.out.printf("#lemmaTagPairs:\t%d%n", lemmaTagCounter.keySet().size());
System.out.printf("#feattags:\t%d%n", reducedTagCounter.keySet().size());
System.out.printf("#feattag+lemmas:\t%d%n", reducedTagLemmaCounter.keySet().size());
System.out.printf("#richtags:\t%d%n", richTagCounter.keySet().size());
System.out.printf("#richtag+lemma:\t%d%n", morphCounter.keySet().size());
System.out.printf("#richtag+lemmaTagPairs:\t%d%n", morphTagCounter.keySet().size());
// Extra
System.out.println("==================");
StringBuilder sbNoLemma = new StringBuilder();
StringBuilder sbMultLemmas = new StringBuilder();
for (Map.Entry<String, Set<String>> wordLemmas : wordLemmaMap.entrySet()) {
String word = wordLemmas.getKey();
Set<String> lemmas = wordLemmas.getValue();
if (lemmas.size() == 0) {
sbNoLemma.append("NO LEMMAS FOR WORD: " + word + "\n");
continue;
}
if (lemmas.size() > 1) {
sbMultLemmas.append("MULTIPLE LEMMAS: " + word + " " + setToString(lemmas) + "\n");
continue;
}
String lemma = lemmas.iterator().next();
Set<String> reducedTags = lemmaReducedTagCounter.getCounter(lemma).keySet();
if (reducedTags.size() > 1) {
System.out.printf("%s --> %s%n", word, lemma);
for (String reducedTag : reducedTags) {
int count = lemmaReducedTagCounter.getCount(lemma, reducedTag);
String posTags =
setToString(reducedTagTagCounter.getCounter(lemma + reducedTag).keySet());
System.out.printf("\t%s\t%d\t%s%n", reducedTag, count, posTags);
}
System.out.println();
}
}
System.out.println("==================");
System.out.println(sbNoLemma.toString());
System.out.println(sbMultLemmas.toString());
System.out.println("==================");
List<String> tags = new ArrayList<>(tagReducedTagCounter.firstKeySet());
Collections.sort(tags);
for (String tag : tags) {
System.out.println(tag);
Set<String> reducedTags = tagReducedTagCounter.getCounter(tag).keySet();
for (String reducedTag : reducedTags) {
int count = tagReducedTagCounter.getCount(tag, reducedTag);
// reducedTag = reducedTag.equals("") ? "NONE" : reducedTag;
System.out.printf("\t%s\t%d%n", reducedTag, count);
}
System.out.println();
}
System.out.println("==================");
}
// when finished = false; break; is called, it means I successfully matched.
@SuppressWarnings("null")
private void goToNextNodeMatch() {
decommitVariableGroups(); // make sure variable groups are free.
decommitNamedNodes();
decommitNamedRelations();
finished = true;
Matcher m = null;
while (nodeMatchCandidateIterator.hasNext()) {
if (myNode.reln.getName() != null) {
String foundReln = namesToRelations.get(myNode.reln.getName());
nextMatchReln = ((GraphRelation.SearchNodeIterator) nodeMatchCandidateIterator).getReln();
if ((foundReln != null) && (!nextMatchReln.equals(foundReln))) {
nextMatch = nodeMatchCandidateIterator.next();
continue;
}
}
nextMatch = nodeMatchCandidateIterator.next();
// System.err.println("going to next match: " + nextMatch.word() + " " +
// myNode.descString + " " + myNode.isLink);
if (myNode.descString.equals("{}") && myNode.isLink) {
IndexedWord otherNode = namesToNodes.get(myNode.name);
if (otherNode != null) {
if (otherNode.equals(nextMatch)) {
if (!myNode.negDesc) {
finished = false;
break;
}
} else {
if (myNode.negDesc) {
finished = false;
break;
}
}
} else {
boolean found = myNode.nodeAttrMatch(nextMatch, hyp ? sg : sg_aligned, ignoreCase);
if (found) {
for (Pair<Integer, String> varGroup : myNode.variableGroups) {
// if variables have been captured from a regex, they
// must match any previous matchings
String thisVariable = varGroup.second();
String thisVarString = variableStrings.getString(thisVariable);
if (thisVarString != null && !thisVarString.equals(m.group(varGroup.first()))) {
// failed to match a variable
found = false;
break;
}
}
// nodeAttrMatch already checks negDesc, so no need to
// check for that here
finished = false;
break;
}
}
} else { // try to match the description pattern.
boolean found = myNode.nodeAttrMatch(nextMatch, hyp ? sg : sg_aligned, ignoreCase);
if (found) {
for (Pair<Integer, String> varGroup : myNode.variableGroups) {
// if variables have been captured from a regex, they
// must match any previous matchings
String thisVariable = varGroup.second();
String thisVarString = variableStrings.getString(thisVariable);
if (thisVarString != null && !thisVarString.equals(m.group(varGroup.first()))) {
// failed to match a variable
found = false;
break;
}
}
// nodeAttrMatch already checks negDesc, so no need to
// check for that here
finished = false;
break;
}
}
} // end while
if (!finished) { // I successfully matched.
resetChild();
if (myNode.name != null) {
// note: have to fill in the map as we go for backreferencing
if (!namesToNodes.containsKey(myNode.name)) {
// System.err.println("making namedFirst");
namedFirst = true;
}
// System.err.println("adding named node: " + myNode.name + "=" +
// nextMatch.word());
namesToNodes.put(myNode.name, nextMatch);
}
if (myNode.reln.getName() != null) {
if (!namesToRelations.containsKey(myNode.reln.getName())) relnNamedFirst = true;
namesToRelations.put(myNode.reln.getName(), nextMatchReln);
}
commitVariableGroups(m); // commit my variable groups.
}
// finished is false exiting this if and only if nextChild exists
// and has a label or backreference that matches
// (also it will just have been reset)
}
/**
* Usage: java edu.stanford.nlp.trees.tregex.tsurgeon.Tsurgeon [-s] -treeFile file-with-trees [-po
* matching-pattern operation] operation-file-1 operation-file-2 ... operation-file-n
*
* <h4>Arguments:</h4>
*
* Each argument should be the name of a transformation file that contains a list of pattern and
* transformation operation list pairs. That is, it is a sequence of pairs of a {@link
* TregexPattern} pattern on one or more lines, then a blank line (empty or whitespace), then a
* list of transformation operations one per line (as specified by <b>Legal operation syntax</b>
* below) to apply when the pattern is matched, and then another blank line (empty or whitespace).
* Note the need for blank lines: The code crashes if they are not present as separators (although
* the blank line at the end of the file can be omitted). The script file can include comment
* lines, either whole comment lines or trailing comments introduced by %, which extend to the end
* of line. A needed percent mark can be escaped by a preceding backslash.
*
* <p>For example, if you want to excise an SBARQ node whenever it is the parent of an SQ node,
* and relabel the SQ node to S, your transformation file would look like this:
*
* <blockquote>
*
* <code>
* SBARQ=n1 < SQ=n2<br>
* <br>
* excise n1 n1<br>
* relabel n2 S
* </code>
*
* </blockquote>
*
* <p>
*
* <h4>Options:</h4>
*
* <ul>
* <li><code>-treeFile <filename></code> specify the name of the file that has the trees
* you want to transform.
* <li><code>-po <matchPattern> <operation></code> Apply a single operation to
* every tree using the specified match pattern and the specified operation. Use this option
* when you want to quickly try the effect of one pattern/surgery combination, and are too
* lazy to write a transformation file.
* <li><code>-s</code> Print each output tree on one line (default is pretty-printing).
* <li><code>-m</code> For every tree that had a matching pattern, print "before" (prepended as
* "Operated on:") and "after" (prepended as "Result:"). Unoperated trees just pass through
* the transducer as usual.
* <li><code>-encoding X</code> Uses character set X for input and output of trees.
* <li><code>-macros <filename></code> A file of macros to use on the tregex pattern.
* Macros should be one per line, with original and replacement separated by tabs.
* <li><code>-hf <headfinder-class-name></code> use the specified {@link HeadFinder} class
* to determine headship relations.
* <li><code>-hfArg <string></code> pass a string argument in to the {@link HeadFinder}
* class's constructor. <code>-hfArg</code> can be used multiple times to pass in multiple
* arguments.
* <li><code>-trf <TreeReaderFactory-class-name></code> use the specified {@link
* TreeReaderFactory} class to read trees from files.
* </ul>
*
* <h4>Legal operation syntax:</h4>
*
* <ul>
* <li><code>delete <name></code> deletes the node and everything below it.
* <li><code>prune <name></code> Like delete, but if, after the pruning, the parent has
* no children anymore, the parent is pruned too. Pruning continues to affect all ancestors
* until one is found with remaining children. This may result in a null tree.
* <li><code>excise <name1> <name2></code> The name1 node should either dominate
* or be the same as the name2 node. This excises out everything from name1 to name2. All
* the children of name2 go into the parent of name1, where name1 was.
* <li><code>relabel <name> <new-label></code> Relabels the node to have the new
* label. <br>
* There are three possible forms: <br>
* <code>relabel nodeX VP</code> - for changing a node label to an alphanumeric string <br>
* <code>relabel nodeX /''/</code> - for relabeling a node to something that isn't a valid
* identifier without quoting <br>
* <code>relabel nodeX /^VB(.*)$/verb\\/$1/</code> - for regular expression based
* relabeling. In this case, all matches of the regular expression against the node label
* are replaced with the replacement String. This has the semantics of Java/Perl's
* replaceAll: you may use capturing groups and put them in replacements with $n. For
* example, if the pattern is /foo/bar/ and the node matched is "foo", the replaceAll
* semantics result in "barbar". If the pattern is /^foo(.*)$/bar$1/ and node matched is
* "foofoo", relabel will result in "barfoo". <br>
* When using the regex replacement method, you can also use the sequences ={node} and
* %{var} in the replacement string to use captured nodes or variable strings in the
* replacement string. For example, if the Tregex pattern was "duck=bar" and the relabel is
* /foo/={bar}/, "foofoo" will be replaced with "duckduck". <br>
* To concatenate two nodes named in the tregex pattern, for example, you can use the
* pattern /^.*$/={foo}={bar}/. Note that the ^.*$ is necessary to make sure the regex
* pattern only matches and replaces once on the entire node name. <br>
* To get an "=" or a "%" in the replacement, using \ escaping. Also, as in the example you
* can escape a slash in the middle of the second and third forms with \\/ and \\\\. <br>
* <li><code>insert <name> <position></code> or <code>
* insert <tree> <position></code> inserts the named node or tree into the
* position specified.
* <li><code>move <name> <position></code> moves the named node into the
* specified position.
* <p>Right now the only ways to specify position are:
* <p><code>$+ <name></code> the left sister of the named node<br>
* <code>$- <name></code> the right sister of the named node<br>
* <code>>i <name></code> the i_th daughter of the named node<br>
* <code>>-i <name></code> the i_th daughter, counting from the right, of the
* named node.
* <li><code>replace <name1> <name2></code> deletes name1 and inserts a copy of
* name2 in its place.
* <li><code>replace <name> <tree> <tree2>...</code> deletes name and
* inserts the new tree(s) in its place. If more than one replacement tree is given, each of
* the new subtrees will be added in order where the old tree was. Multiple subtrees at the
* root is an illegal operation and will throw an exception.
* <li>{@code createSubtree <new-label> <name1> [<name2>]} Create a subtree out of all the nodes
* from {@code <name1>} through {@code <name2>} and puts the new subtree where that span
* used to be. To limit the operation to just one node, elide {@code <name2>}.
* <li><code>adjoin <auxiliary_tree> <name></code> Adjoins the specified auxiliary
* tree into the named node. The daughters of the target node will become the daughters of
* the foot of the auxiliary tree.
* <li><code>adjoinH <auxiliary_tree> <name></code> Similar to adjoin, but
* preserves the target node and makes it the root of <tree>. (It is still accessible
* as <code>name</code>. The root of the auxiliary tree is ignored.)
* <li><code>adjoinF <auxiliary_tree> <name></code> Similar to adjoin, but
* preserves the target node and makes it the foot of <tree>. (It is still accessible
* as <code>name</code>, and retains its status as parent of its children. The root of the
* auxiliary tree is ignored.)
* <li>
* <dt><code>coindex <name1> <name2> ... <nameM> </code> Puts a (Penn
* Treebank style) coindexation suffix of the form "-N" on each of nodes name_1 through
* name_m. The value of N will be automatically generated in reference to the existing
* coindexations in the tree, so that there is never an accidental clash of indices across
* things that are not meant to be coindexed.
* </ul>
*
* <p>In the context of <code>adjoin</code>, <code>adjoinH</code>, and <code>adjoinF</code>, an
* auxiliary tree is a tree in Penn Treebank format with <code>@</code> on exactly one of the
* leaves denoting the foot of the tree. The operations which use the foot use the labeled node.
* For example: <br>
* Tsurgeon: <code>adjoin (FOO (BAR@)) foo</code> <br>
* Tregex: <code>B=foo</code> <br>
* Input: <code>(A (B 1 2))</code> Output: <code>(A (FOO (BAR 1 2)))</code>
*
* <p>Tsurgeon applies the same operation to the same tree for as long as the given tregex
* operation matches. This means that infinite loops are very easy to cause. One common situation
* where this comes up is with an insert operation will repeats infinitely many times unless you
* add an expression to the tregex that matches against the inserted pattern. For example, this
* pattern will infinite loop:
*
* <blockquote>
*
* <code>
* TregexPattern tregex = TregexPattern.compile("S=node << NP"); <br>
* TsurgeonPattern tsurgeon = Tsurgeon.parseOperation("insert (NP foo) >-1 node");
* </code>
*
* </blockquote>
*
* This pattern, though, will terminate:
*
* <blockquote>
*
* <code>
* TregexPattern tregex = TregexPattern.compile("S=node << NP !<< foo"); <br>
* TsurgeonPattern tsurgeon = Tsurgeon.parseOperation("insert (NP foo) >-1 node");
* </code>
*
* </blockquote>
*
* <p>Tsurgeon has (very) limited support for conditional statements. If a pattern is prefaced
* with <code>if exists <name></code>, the rest of the pattern will only execute if the
* named node was found in the corresponding TregexMatcher.
*
* @param args a list of names of files each of which contains a single tregex matching pattern
* plus a list, one per line, of transformation operations to apply to the matched pattern.
* @throws Exception If an I/O or pattern syntax error
*/
public static void main(String[] args) throws Exception {
String headFinderClassName = null;
String headFinderOption = "-hf";
String[] headFinderArgs = null;
String headFinderArgOption = "-hfArg";
String encoding = "UTF-8";
String encodingOption = "-encoding";
if (args.length == 0) {
System.err.println(
"Usage: java edu.stanford.nlp.trees.tregex.tsurgeon.Tsurgeon [-s] -treeFile <file-with-trees> [-po <matching-pattern> <operation>] <operation-file-1> <operation-file-2> ... <operation-file-n>");
System.exit(0);
}
String treePrintFormats;
String singleLineOption = "-s";
String verboseOption = "-v";
String matchedOption =
"-m"; // if set, then print original form of trees that are matched & thus operated on
String patternOperationOption = "-po";
String treeFileOption = "-treeFile";
String trfOption = "-trf";
String macroOption = "-macros";
String macroFilename = "";
Map<String, Integer> flagMap = Generics.newHashMap();
flagMap.put(patternOperationOption, 2);
flagMap.put(treeFileOption, 1);
flagMap.put(trfOption, 1);
flagMap.put(singleLineOption, 0);
flagMap.put(encodingOption, 1);
flagMap.put(headFinderOption, 1);
flagMap.put(macroOption, 1);
Map<String, String[]> argsMap = StringUtils.argsToMap(args, flagMap);
args = argsMap.get(null);
if (argsMap.containsKey(headFinderOption))
headFinderClassName = argsMap.get(headFinderOption)[0];
if (argsMap.containsKey(headFinderArgOption)) headFinderArgs = argsMap.get(headFinderArgOption);
if (argsMap.containsKey(verboseOption)) verbose = true;
if (argsMap.containsKey(singleLineOption)) treePrintFormats = "oneline,";
else treePrintFormats = "penn,";
if (argsMap.containsKey(encodingOption)) encoding = argsMap.get(encodingOption)[0];
if (argsMap.containsKey(macroOption)) macroFilename = argsMap.get(macroOption)[0];
TreePrint tp = new TreePrint(treePrintFormats, new PennTreebankLanguagePack());
PrintWriter pwOut = new PrintWriter(new OutputStreamWriter(System.out, encoding), true);
TreeReaderFactory trf;
if (argsMap.containsKey(trfOption)) {
String trfClass = argsMap.get(trfOption)[0];
trf = ReflectionLoading.loadByReflection(trfClass);
} else {
trf = new TregexPattern.TRegexTreeReaderFactory();
}
Treebank trees = new DiskTreebank(trf, encoding);
if (argsMap.containsKey(treeFileOption)) {
trees.loadPath(argsMap.get(treeFileOption)[0]);
}
List<Pair<TregexPattern, TsurgeonPattern>> ops =
new ArrayList<Pair<TregexPattern, TsurgeonPattern>>();
TregexPatternCompiler compiler;
if (headFinderClassName == null) {
compiler = new TregexPatternCompiler();
} else {
HeadFinder hf;
if (headFinderArgs == null) {
hf = ReflectionLoading.loadByReflection(headFinderClassName);
} else {
hf = ReflectionLoading.loadByReflection(headFinderClassName, (Object[]) headFinderArgs);
}
compiler = new TregexPatternCompiler(hf);
}
Macros.addAllMacros(compiler, macroFilename, encoding);
if (argsMap.containsKey(patternOperationOption)) {
TregexPattern matchPattern = compiler.compile(argsMap.get(patternOperationOption)[0]);
TsurgeonPattern p = parseOperation(argsMap.get(patternOperationOption)[1]);
ops.add(new Pair<TregexPattern, TsurgeonPattern>(matchPattern, p));
} else {
for (String arg : args) {
List<Pair<TregexPattern, TsurgeonPattern>> pairs =
getOperationsFromFile(arg, encoding, compiler);
for (Pair<TregexPattern, TsurgeonPattern> pair : pairs) {
if (verbose) {
System.err.println(pair.second());
}
ops.add(pair);
}
}
}
for (Tree t : trees) {
Tree original = t.deepCopy();
Tree result = processPatternsOnTree(ops, t);
if (argsMap.containsKey(matchedOption) && matchedOnTree) {
pwOut.println("Operated on: ");
displayTree(original, tp, pwOut);
pwOut.println("Result: ");
}
displayTree(result, tp, pwOut);
}
}
private void incrementDay(ISODateInstance referenceDate, Pair<DateField, Integer> relation) {
String origDateString = referenceDate.getStartDate();
String dayString =
origDateString.substring(origDateString.length() - 2, origDateString.length());
if (dayString.contains("*")) {
isoDate = origDateString;
return;
}
// Date is not a variable
Integer dayNum = Integer.parseInt(dayString);
String monthString =
origDateString.substring(origDateString.length() - 4, origDateString.length() - 2);
int numDaysInMonth = 30; // default - assume this if month is a variable
int monthNum =
-1; // ie, we don't know the month yet - this remains -1 if the month is a variable
if (!monthString.contains("*")) {
// Set appropriate numDaysInMonth and monthNum
monthNum = Integer.parseInt(monthString);
numDaysInMonth = daysPerMonth.get(monthNum);
}
// Now, find out if we're an edge case (potential to increment month)
if (dayNum + relation.second() <= numDaysInMonth && dayNum + relation.second() >= 1) {
// Not an edge case - just increment the day, create a new string, and return
dayNum += relation.second();
isoDate = makeStringDayChange(origDateString, dayNum);
return;
}
// Since we're an edge case, the month can't be a variable - if it is a variable, just set this
// to the reference string
if (monthNum == -1) {
isoDate = origDateString;
return;
}
// At this point, neither our day nor our month is a variable
isoDate = origDateString;
boolean decreasing = (dayNum + relation.second() < 1);
// Need to increment the month, set the date appropriately - we need the new month num to set
// the day appropriately, so do month first
int newMonthNum;
// Now, check if we're an edge case for month
if ((monthNum + 1 > 12 && !decreasing) || (monthNum - 1 < 1 && decreasing)) {
// First, change the month
if (decreasing) {
newMonthNum = 12;
} else {
newMonthNum = 1;
}
// If we can, increment the year
// TODO: fix this to work more nicely with variables and thus handle more cases
String yearString = origDateString.substring(0, 4);
if (!yearString.contains("*")) {
if (decreasing) {
isoDate = makeStringYearChange(isoDate, Integer.parseInt(yearString) - 1);
} else {
isoDate = makeStringYearChange(isoDate, Integer.parseInt(yearString) + 1);
}
}
} else {
// We're not an edge case for month - just increment
if (decreasing) {
newMonthNum = monthNum - 1;
} else {
newMonthNum = monthNum + 1;
}
}
// do the increment
isoDate = makeStringMonthChange(isoDate, newMonthNum);
int newDateNum;
if (decreasing) {
newDateNum = -relation.second() + daysPerMonth.get(newMonthNum) - dayNum;
} else {
newDateNum = relation.second() - dayNum + daysPerMonth.get(monthNum);
}
// Now, change the day in our original string to be appropriate
isoDate = makeStringDayChange(isoDate, newDateNum);
}
