/**
* Parses a sentence and returns the parse tree.
*
* @param sentence a sentence
* @return Tree character offsets in keys BEGIN_KEY and END_KEY
*/
@SuppressWarnings("unchecked")
public static Tree parseTree(String sentence) {
if (tlp == null || parser == null)
throw new RuntimeException("Parser has not been initialized");
// parse the sentence to produce stanford Tree
log.debug("Parsing sentence");
Tree tree = null;
synchronized (parser) {
Tokenizer tokenizer = tlp.getTokenizerFactory().getTokenizer(new StringReader(sentence));
List<Word> words = tokenizer.tokenize();
log.debug("Tokenization: " + words);
parser.parse(new Sentence(words));
tree = parser.getBestParse();
}
// label tree with character extents
// log.debug("Setting character extents");
// updateTreeLabels(tree, tree, new MutableInteger(), new MutableInteger(-1));
// log.debug("Creating offset mapping");
// List<RangeMap> mapping = createMapping(sentence);
// log.debug(mapping.toString());
// log.debug("Applying offset mapping");
// mapOffsets(tree, mapping);
return tree;
}
// todo: give options for document splitting. A line or the whole file or
// sentence splitting as now
public Iterator<List<IN>> getIterator(Reader r) {
Tokenizer<IN> tokenizer = tokenizerFactory.getTokenizer(r);
// PTBTokenizer.newPTBTokenizer(r, false, true);
List<IN> words = new ArrayList<IN>();
IN previous = tokenFactory.makeToken();
StringBuilder prepend = new StringBuilder();
/*
* This changes SGML tags into whitespace -- it should maybe be moved
* elsewhere
*/
while (tokenizer.hasNext()) {
IN w = tokenizer.next();
String word = w.get(CoreAnnotations.TextAnnotation.class);
Matcher m = sgml.matcher(word);
if (m.matches()) {
String before = StringUtils.getNotNullString(w.get(CoreAnnotations.BeforeAnnotation.class));
String after = StringUtils.getNotNullString(w.get(CoreAnnotations.AfterAnnotation.class));
prepend.append(before).append(word);
String previousTokenAfter =
StringUtils.getNotNullString(previous.get(CoreAnnotations.AfterAnnotation.class));
previous.set(AfterAnnotation.class, previousTokenAfter + word + after);
// previous.appendAfter(w.word() + w.after());
} else {
String before = StringUtils.getNotNullString(w.get(CoreAnnotations.BeforeAnnotation.class));
if (prepend.length() > 0) {
w.set(BeforeAnnotation.class, prepend.toString() + before);
// w.prependBefore(prepend.toString());
prepend = new StringBuilder();
}
words.add(w);
previous = w;
}
}
List<List<IN>> sentences = wts.process(words);
String after = "";
IN last = null;
for (List<IN> sentence : sentences) {
int pos = 0;
for (IN w : sentence) {
w.set(PositionAnnotation.class, Integer.toString(pos));
after = StringUtils.getNotNullString(w.get(CoreAnnotations.AfterAnnotation.class));
w.remove(AfterAnnotation.class);
last = w;
}
}
if (last != null) {
last.set(AfterAnnotation.class, after);
}
return sentences.iterator();
}
@Override
public Sequence<IString> process(String input) {
String tokenizerInput = toUncased(input.trim());
Tokenizer<CoreLabel> tokenizer = tf.getTokenizer(new StringReader(tokenizerInput));
List<String> outputStrings = new LinkedList<>();
while (tokenizer.hasNext()) {
String string = tokenizer.next().get(TextAnnotation.class);
outputStrings.add(string);
}
return IStrings.toIStringSequence(outputStrings);
}
/**
* Parses a sentence and returns the PCFG score as a confidence measure.
*
* @param sentence a sentence
* @return PCFG score
*/
@SuppressWarnings("unchecked")
public static double getPCFGScore(String sentence) {
if (tlp == null || parser == null)
throw new RuntimeException("Parser has not been initialized");
// parse the sentence to produce PCFG score
log.debug("Parsing sentence");
double score;
synchronized (parser) {
Tokenizer tokenizer = tlp.getTokenizerFactory().getTokenizer(new StringReader(sentence));
List<Word> words = tokenizer.tokenize();
log.debug("Tokenization: " + words);
parser.parse(new Sentence(words));
score = parser.getPCFGScore();
}
return score;
}
@Override
public SymmetricalWordAlignment processAndAlign(String input) {
input = input.trim();
// Run through the tokenizer and convert to sequence
String tokenizerInput = toUncased(input);
String[] uncasedInputTokens = tokenizerInput.split("\\s+");
Tokenizer<CoreLabel> tokenizer = tf.getTokenizer(new StringReader(tokenizerInput));
List<CoreLabel> outputTokens = tokenizer.tokenize();
IString[] outputSequence = new IString[outputTokens.size()];
for (int i = 0; i < outputSequence.length; ++i) {
String outputToken = outputTokens.get(i).get(TextAnnotation.class);
outputSequence[i] = new IString(outputToken);
}
// Whitespace tokenization of input, create alignment
Sequence<IString> inputSequence = IStrings.tokenize(input);
assert inputSequence.size() == uncasedInputTokens.length;
SymmetricalWordAlignment alignment =
new SymmetricalWordAlignment(
inputSequence, new SimpleSequence<IString>(true, outputSequence));
// Generate the alignments
StringBuilder inputToken = new StringBuilder();
for (int i = 0, j = 0, limit = outputTokens.size(); j < limit; ++j) {
CoreLabel tokenizedToken = outputTokens.get(j);
String inputTokenPart = toUncased(tokenizedToken.get(OriginalTextAnnotation.class));
alignment.addAlign(i, j);
inputToken.append(inputTokenPart);
if (i >= uncasedInputTokens.length) {
System.err.println("WARNING: Non-invertible input: " + input);
break;
}
if (uncasedInputTokens[i].equals(inputToken.toString())) {
++i;
inputToken = new StringBuilder();
}
}
return alignment;
}
private static List<TypedDependency> getDependencies(String sentence) {
if (pipeline == null) {
loadModels();
}
TokenizerFactory<CoreLabel> tokenizerFactory =
PTBTokenizer.factory(new CoreLabelTokenFactory(), "");
Tokenizer<CoreLabel> tok = tokenizerFactory.getTokenizer(new StringReader(sentence));
List<CoreLabel> rawWords2 = tok.tokenize();
Tree parse = lp.apply(rawWords2);
// parse.pennPrint();
//
// System.out.println(parse.toString());
TreebankLanguagePack tlp = lp.treebankLanguagePack(); // PennTreebankLanguagePack for English
GrammaticalStructureFactory gsf = tlp.grammaticalStructureFactory();
GrammaticalStructure gs = gsf.newGrammaticalStructure(parse);
List<TypedDependency> tdl = gs.typedDependenciesCCprocessed();
return tdl;
}
/**
* parse sentence and generate .trees file
*
* @param en
* @param align
* @param out
*/
public static void parse(String en, String align, String out, boolean verbose) {
// use alignments?
boolean use_alignments = true;
if (align.startsWith("no_align")) {
use_alignments = false;
System.err.println("Not using alignments.");
} else {
System.err.println("Using alignments from " + align);
}
// setup stanfordparser
String grammar = "edu/stanford/nlp/models/lexparser/englishPCFG.ser.gz";
String[] options = {"-outputFormat", "wordsAndTags, typedDependencies"};
LexicalizedParser lp = LexicalizedParser.loadModel(grammar, options);
TreebankLanguagePack tlp = lp.getOp().langpack();
java.util.function.Predicate<java.lang.String> punctuationFilter = x -> true;
GrammaticalStructureFactory gsf =
new edu.stanford.nlp.trees.EnglishGrammaticalStructureFactory(punctuationFilter);
// read document
Iterable<List<? extends HasWord>> sentences;
Reader r = new Reader(en);
String line = null;
List<List<? extends HasWord>> tmp = new ArrayList<List<? extends HasWord>>();
while ((line = r.getNext()) != null) {
Tokenizer<? extends HasWord> token =
tlp.getTokenizerFactory().getTokenizer(new StringReader(line));
List<? extends HasWord> sentence = token.tokenize();
tmp.add(sentence);
}
sentences = tmp;
// set up alignment file reader
Reader alignment = new Reader();
if (use_alignments) {
alignment = new Reader(align);
}
// set up tree file writer
Writer treeWriter = new Writer(out);
// parse
long start = System.currentTimeMillis();
// System.err.print("Parsing sentences ");
int sentID = 0;
for (List<? extends HasWord> sentence : sentences) {
Tree t = new Tree();
// t.setSentID(++sentID);
System.err.println("parse Sentence :" + sentence + "...");
// System.err.print(".");
System.err.println("-----------------------------------------------------------------------");
edu.stanford.nlp.trees.Tree parse = lp.parse(sentence);
// parse.pennPrint();
// List for root node and lexical nodes
List<Node> loneNodes = new LinkedList<Node>();
List<Node> governingNodes = new LinkedList<Node>();
// ROOT node
Node root = new Node(true, true);
root.setTag("ROOT");
t.setRoot(root);
loneNodes.add(root);
governingNodes.add(root);
// tagging
int counter = 0;
String surface = "";
String tag = "";
for (TaggedWord tw : parse.taggedYield()) {
Node n = new Node();
Node governingNode = new Node();
n.setNodeID(++counter);
surface = tw.value();
tag = tw.tag();
if (surface.startsWith("-LRB-")) {
surface = "(";
} else if (surface.startsWith("-RRB-")) {
surface = ")";
// } else if (surface.startsWith("-LSB-")){
// surface = "[";
// } else if (surface.startsWith("-RSB-")){
// surface = "]";
// } else if (surface.startsWith("-LCB-")){
// surface = "{";
// } else if (surface.startsWith("-RCB-")){
// surface = "}";
} else if (surface.startsWith("''")) {
surface = "\"";
}
tag = tag.replaceAll("#", "-NUM-");
surface = surface.replaceAll("&", "-AMP-");
surface = surface.replaceAll("#", "-NUM-");
surface = surface.replaceAll(">", "-GRE-");
surface = surface.replaceAll("=", "-EQU-");
n.setInitialLexicalIndex(counter);
governingNode.setInitialLexicalIndex(counter);
n.setSurface(surface);
// System.out.print("("+tw.value()+" : ");
n.setTag(tag);
governingNode.setTag("_" + tag);
governingNode.setLabel("_gov");
// System.out.print(tw.tag()+")");
loneNodes.add(n);
governingNodes.add(governingNode);
governingNode.setChild(n);
}
// System.out.println("");
// t.setSentLength(t.getNodes().size() - 1);
// List<Node> loneNodes = new LinkedList<Node>();
Node[] nodes = new Node[2000];
// labeling
int depIndex;
int govIndex;
String[] depInfo;
String[] govInfo;
GrammaticalStructure gs = gsf.newGrammaticalStructure(parse);
List<TypedDependency> tdl = gs.typedDependencies(false);
// List<TypedDependency> tdl = gs.typedDependenciesCCprocessed();
for (TypedDependency td : tdl) {
depIndex = td.dep().index();
govIndex = td.gov().index();
// System.out.println("Index1:"+depIndex);
// System.out.println("Index2:"+govIndex);
// if (nodes[depIndex] == null){
// System.out.println("Making node!");
// nodes[depIndex] = new Node();
// }
// if (nodes[govIndex] == null){
// System.out.println("Making node!");
// nodes[govIndex] = new Node();
// }
Node dep = loneNodes.get((depIndex));
Node gov = governingNodes.get((govIndex));
Node depcopy = governingNodes.get((depIndex));
Node govcopy = loneNodes.get((govIndex));
dep.setLabel(td.reln().toString());
depcopy.setLabel(td.reln().toString());
govcopy.setLabel("head");
// System.out.println(td.toString());
govInfo = td.gov().toString().split("/");
depInfo = td.dep().toString().split("/");
// System.out.println(td.gov().toString());
// System.out.println(td.dep().toString());
// dep.setSurface(depInfo[0]);
// dep.setTag(depInfo[1]);
gov.setChild(governingNodes.get(depIndex));
governingNodes.get(depIndex).setParent(gov);
// gov.setChild(dep);
dep.setParent(governingNodes.get(depIndex));
}
// t.setRoot(nodes[0]);
// Collapse tree to remove unneeded governing nodes:
Node gov;
Node dep;
Node parent;
List<Node> children;
for (int i = 1; i < governingNodes.size(); i++) { // start with index 1 to skip root
gov = governingNodes.get(i);
dep = loneNodes.get(i);
if (gov.getChildren().size() <= 1) {
int k = 0;
parent = gov.getParent();
children = parent.getChildren();
for (Node n : children) {
if (n == gov) {
gov.getParent().replaceChild(k, dep);
dep.setParent(gov.getParent());
}
k++;
}
}
}
// Mark head nodes with appropriate label:
int k = 0;
for (Node n : loneNodes) {
if (k != 0) {
if (n.getLabel() == n.getParent().getLabel()) {
n.setLabel("head");
}
} else {
n.setLabel("null");
}
k++;
}
// Sort lexical children of each governing node in lexical order
for (Node n : governingNodes) {
n.sortChildrenByInitialIndex();
}
// combine with alignment
if (use_alignments) {
t.initialize(alignment.readNextAlign());
} else {
t.initializeUnaligned();
}
// write tree to file
treeWriter.write(t);
// print tree to console
System.out.println(t.toSentence());
if (verbose) {
System.err.println(t.toString());
// t.recursivePrint();
}
System.err.println("#######################################################################");
}
long stop = System.currentTimeMillis();
System.err.println("...done! [" + (stop - start) / 1000 + " sec].");
treeWriter.close();
}
private Tree getPosTree(String sentence) {
final Tokenizer<CoreLabel> tokenizer =
tokenizerFactory.getTokenizer(new StringReader(sentence));
final List<CoreLabel> tokens = tokenizer.tokenize();
return parser.apply(tokens);
}
private static List<CoreLabel> tokenize(String str) {
Tokenizer<CoreLabel> tokenizer = tokenizerFactory.getTokenizer(new StringReader(str));
return tokenizer.tokenize();
}
public HashMap<String,ArrayList<TreeData>> parseAllDocs() throws IOException{
String grammar = "./jsan_resources/englishPCFG.ser.gz";
String[] options = { "-maxLength", "120", "-retainTmpSubcategories" };
// LexicalizedParser lp = new LexicalizedParser(grammar, options);
LexicalizedParser lp = new LexicalizedParser()
TreebankLanguagePack tlp = new PennTreebankLanguagePack();
GrammaticalStructureFactory gsf = tlp.grammaticalStructureFactory();
Iterable<List<? extends HasWord>> sentences;
ArrayList<HashMap<String,ArrayList<String>>> everything = new ArrayList<HashMap<String,ArrayList<String>>>(3);
everything.add(0,otherSampleStrings);
everything.add(1,authorSampleStrings);
everything.add(2,toModifyStrings);
Iterator<HashMap<String,ArrayList<String>>> everythingIter = everything.iterator();
int docTypeNumber = -1; // 0 for otherSampleStrings, 1 for authorSampleStrings, 2 for toModifyStrings
int numLoaded = 0;
while(everythingIter.hasNext()){
docTypeNumber++;
HashMap<String,ArrayList<String>> currentSampleStrings = docPathFinder();
Set<String> currentDocStrings = currentSampleStrings.keySet();
Iterator<String> docStrIter = currentDocStrings.iterator();
String docID;
ArrayList<String> sentenceTokens;
allTreeProcessors[docTypeNumber] = new TreeProcessor();
allTreeProcessors[docTypeNumber].clearLoadedTreeDataMaps();
numLoaded=0;
while(docStrIter.hasNext()){
docID = docStrIter.next();
sentenceTokens = currentSampleStrings.get(docID);
if(sentenceTokens == null){
allTreeProcessors[docTypeNumber].loadTreeDataMap(docID, GRAMMAR_DIR, false);
numLoaded++;
continue;
}
//System.out.println(sentenceTokens.size()+", strIter.hasNext? -> "+strIter.hasNext());
numSentences = sentenceTokens.size();
//initialize(numSentences);
Iterator<String> sentIter = sentenceTokens.iterator();
List<List<? extends HasWord>> tmp = new ArrayList<List<? extends HasWord>>();
String tempSent;
while(sentIter.hasNext()){
tempSent = sentIter.next();
Tokenizer<? extends HasWord> toke = tlp.getTokenizerFactory().getTokenizer(new StringReader(tempSent));
List<? extends HasWord> sentenceTokenized = toke.tokenize();
tmp.add(sentenceTokenized);
}
sentences = tmp;
//int numDone = 0;
TreeProcessor.singleDocMap.clear();
boolean willSaveResults = true;
for (List<? extends HasWord> sentence : sentences) {
Tree parse = lp.apply(sentence);
//parse.pennPrint();
//System.out.println(parse.treeSkeletonCopy().toString());
//System.out.println(parse.taggedYield());
//System.out.println();
//printSubTrees(parse);
//TreeContainer.recurseTree(parse,"breadth");
allTreeProcessors[docTypeNumber].processTree(parse, 0, willSaveResults);
//System.out.println(tc.processedTrees.toString().replaceAll("\\]\\], \\(","\\]\\]\n\\("));
//numDone++;
//System.out.println("sent "+numDone+" of "+numSentences+" done ");
//System.out.println(tc.processedTrees.toString());
//in.nextLine();
//TreeContainer.recurseTree(parse, "depth");
//in.nextLine();
//addTree(parse);
//GrammaticalStructure gs = gsf.newGrammaticalStructure(parse);//TODO: LOOK AT THIS
//Collection tdl = gs.typedDependenciesCCprocessed(true);
//System.out.println(tdl);
//System.out.println();
}
if(willSaveResults == true)
ObjectIO.writeObject(TreeProcessor.singleDocMap,docID, GRAMMAR_DIR);
//System.out.println("After all sents: ");
//System.out.println(tc.processedTrees.toString().replaceAll("\\]\\], \\(","\\]\\]\n\\("));
//String sent3 = "This is one last test!";
//Tree parse3 = lp.apply(sent3);
//parse3.pennPrint();
//System.out.println("After sorting and writing:");
//System.out.println(tc.processedTrees.toString().replaceAll("\\]\\], \\(","\\]\\]\n\\("));
//Scanner in = new Scanner(System.in);
//System.out.println("First one done.");
//in.nextLine();
//viewTrees();
}
//TreeProcessor.writeTreeDataToCSV(sortedTD,docID);
allTreeProcessors[docTypeNumber].unmergedMaps = new ArrayList<HashMap<String,TreeData>>(numLoaded+1);
}
int i= 0;
allParsedAndOrdered.clear();
String[] docTypes = new String[]{"otherSample","authorSample","toModify"};
for(i=0; i < 3; i++){
allTreeProcessors[i].unmergedMaps.add(allTreeProcessors[i].processedTrees);
allTreeProcessors[i].unmergedMaps.addAll(allTreeProcessors[i].loadedTreeDataMaps);
allTreeProcessors[i].mergeTreeDataLists(allTreeProcessors[i].unmergedMaps);
allParsedAndOrdered.put(docTypes[i],allTreeProcessors[i].sortTreeData(allTreeProcessors[i].mergedMap));
}
//ArrayList<TreeData> sortedTD = TreeContainer.sortTreeData(TreeContainer.allProcessedTrees);
//TreeContainer.writeTreeDataToCSV(sortedTD,"ALL_AUTHORS");
return allParsedAndOrdered;
}
/**
* Tokenize a sentence in the argument, and print out the tokens to the console.
*
* @param args Set the first argument as the sentence to
* <p>be tokenized.
*/
public static void main(String[] args) {
TokenizerFactory<Word> factory = PTBTokenizerFactory.newTokenizerFactory();
Tokenizer<Word> tokenizer = factory.getTokenizer(new StringReader(args[0]));
System.out.println(tokenizer.tokenize());
}
public static void main(String[] args) throws IOException {
Tokenizer<String> att = new ArabicTreebankTokenizer(new FileReader(args[0]));
while (att.hasNext()) {
System.out.print(att.next());
}
}