/** splits document 'doc' into sentences, adding 'sentence' annotations */
static void addSentences(Document doc) {
SpecialZoner.findSpecialZones(doc);
Vector<Annotation> textSegments = doc.annotationsOfType("TEXT");
if (textSegments == null) {
System.out.println("No <TEXT> in document");
return;
}
for (Annotation ann : textSegments) {
Span textSpan = ann.span();
// check document case
Ace.monocase = Ace.allLowerCase(doc);
// split into sentences
SentenceSplitter.split(doc, textSpan);
}
Vector<Annotation> sentences = doc.annotationsOfType("sentence");
if (sentences != null) {
int sentNo = 0;
for (Annotation sentence : sentences) {
sentNo++;
sentence.put("ID", "SENT-" + sentNo);
}
}
doc.removeAnnotationsOfType("dateline");
doc.removeAnnotationsOfType("textBreak");
doc.shrink("sentence");
}
/**
* hides (adds the 'hidden' feature) to all annotations of type <I>type</I> beginning at the
* starting position of span <I>span</I>.
*/
public static void hideAnnotations(Document doc, String type, Span span) {
for (int posn = span.start(); posn < span.end(); posn++) {
Vector annotations = doc.annotationsAt(posn, type);
if (annotations != null) {
for (int i = 0; i < annotations.size(); i++) {
Annotation ann = (Annotation) annotations.elementAt(i);
ann.put("hidden", "true");
// Console.println ("Hiding " + ann);
}
}
}
}
static void writeDoc1(Document doc, PrintStream out) throws IOException {
Vector<Annotation> entities = doc.annotationsOfType("entity");
if (entities == null) {
System.err.println("No Entity: " + doc);
return;
}
Iterator<Annotation> entityIt = entities.iterator();
int i = 0;
while (entityIt.hasNext()) {
Annotation entity = entityIt.next();
Vector mentions = (Vector) entity.get("mentions");
Iterator mentionIt = mentions.iterator();
String nameType = (String) entity.get("nameType");
while (mentionIt.hasNext()) {
Annotation mention1 = (Annotation) mentionIt.next();
Annotation mention2 = new Annotation("refobj", mention1.span(), new FeatureSet());
mention2.put("objid", Integer.toString(i));
if (nameType != null) {
mention2.put("netype", nameType);
}
doc.addAnnotation(mention2);
}
i++;
}
// remove other annotations.
String[] annotypes = doc.getAnnotationTypes();
for (i = 0; i < annotypes.length; i++) {
String t = annotypes[i];
if (!(t.equals("tagger") || t.equals("refobj") || t.equals("ENAMEX"))) {
doc.removeAnnotationsOfType(t);
}
}
writeDocRaw(doc, out);
return;
}
/**
* parse all the sentences in Document 'doc', returning a SyntacticRelationSet containing all the
* dependency relations.
*/
public static SyntacticRelationSet parseDocument(Document doc) {
Vector<Annotation> sentences = doc.annotationsOfType("sentence");
if (sentences == null || sentences.size() == 0) {
System.out.println("DepParser: no sentences");
return null;
}
if (fsw == null) {
System.out.println("DepParser: no model loaded");
return null;
}
SyntacticRelationSet relations = new SyntacticRelationSet();
for (Annotation sentence : sentences) {
Span span = sentence.span();
parseSentence(doc, span, relations);
}
return relations;
}
/** generate the dependency parse for a sentence, adding its arcs to 'relations'. */
public static void parseSentence(Document doc, Span span, SyntacticRelationSet relations) {
if (fsw == null) {
System.out.println("DepParser: no model loaded");
return;
}
// System.out.println ("parseSentence: " + doc.text(span));
// run Penn part-of-speech tagger
// JetTest.tagger.annotate(doc, span, "tagger");
// build sentence
List<Token> tokens = new ArrayList<Token>();
List<Integer> offset = new ArrayList<Integer>();
offset.add(0); // don't use 0th entry
int tokenNum = 0;
int posn = span.start();
while (posn < span.end()) {
tokenNum++;
Annotation tokenAnnotation = doc.tokenAt(posn);
for (String s : SPECIAL_TOKEN) {
Vector<Annotation> va = doc.annotationsAt(posn, s);
if (va != null && va.size() > 0) {
tokenAnnotation = va.get(0);
break;
}
}
if (tokenAnnotation == null) return;
String tokenText = doc.normalizedText(tokenAnnotation).replaceAll(" ", "_");
Vector v = doc.annotationsAt(posn, "tagger");
Annotation a = (Annotation) v.get(0);
String pos = (String) a.get("cat");
tokens.add(new Token(tokenText, pos, tokenNum));
offset.add(posn);
if (posn >= tokenAnnotation.end()) {
break;
}
posn = tokenAnnotation.end();
}
Sentence sent = new Sentence(tokens);
// parse sentence
Arc[] arcs =
fsw.process(
sent,
tokens.size() > 0 && tokens.get(0).getPos() == null,
true,
true,
true,
true,
true)
.getParse()
.getHeadArcs();
// get dependencies
for (Arc arc : arcs) {
if (arc == null) continue;
if (arc.getDependency().equalsIgnoreCase("ROOT")) continue;
Token head = arc.getHead();
String headText = head.getText();
String headPos = head.getPos();
Integer headOffset = offset.get(head.getIndex());
Token dep = arc.getChild();
String depText = dep.getText();
String depPos = dep.getPos();
Integer depOffset = offset.get(dep.getIndex());
String type = arc.getDependency();
SyntacticRelation r =
new SyntacticRelation(headOffset, headText, headPos, type, depOffset, depText, depPos);
relations.add(r);
// System.out.println ("parseSentence: adding relation " + r);
}
}