/** generate mention annotations (with entity numbers) based on the ACE entities and mentions. */
static void addMentionTags(Document doc, AceDocument aceDoc) {
ArrayList<AceEntity> entities = aceDoc.entities;
for (int i = 0; i < entities.size(); i++) {
AceEntity entity = entities.get(i);
ArrayList<AceEntityMention> mentions = entity.mentions;
for (int j = 0; j < mentions.size(); j++) {
AceEntityMention mention = (AceEntityMention) mentions.get(j);
// we compute a jetSpan not including trailing whitespace
Span aceSpan = mention.head;
// skip mentions in ChEnglish APF not aligned to any English text
if (aceSpan.start() < 0) continue;
Span jetSpan = new Span(aceSpan.start(), aceSpan.end() + 1);
FeatureSet features = new FeatureSet("entity", new Integer(i));
if (flags.contains("types")) {
features.put("type", entity.type.substring(0, 3));
if (entity.subtype != null) features.put("subtype", entity.subtype);
}
if (flags.contains("extents")) {
String cleanExtent = mention.text.replaceAll("\n", " ");
features.put("extent", AceEntityMention.addXmlEscapes(cleanExtent));
}
doc.annotate("mention", jetSpan, features);
}
}
}
static void addTimexTags(Document doc, AceDocument aceDoc) {
List<AceTimex> timeExpressions = aceDoc.timeExpressions;
for (AceTimex timex : timeExpressions) {
AceTimexMention mention = (AceTimexMention) timex.mentions.get(0);
Span aceSpan = mention.extent;
Span jetSpan = new Span(aceSpan.start(), aceSpan.end() + 1);
FeatureSet features = new FeatureSet();
if (timex.val != null && !timex.val.equals("")) features.put("val", timex.val);
if (timex.anchorVal != null && !timex.anchorVal.equals(""))
features.put("anchor_val", timex.anchorVal);
if (timex.anchorDir != null && !timex.anchorDir.equals(""))
features.put("anchor_dir", timex.anchorDir);
if (timex.set != null && !timex.set.equals("")) features.put("set", timex.set);
if (timex.mod != null && !timex.mod.equals("")) features.put("mod", timex.mod);
doc.annotate("timex2", jetSpan, features);
}
}
/** generate mention annotations (with entity numbers) based on the ACE entities and mentions. */
static void addMentionTags(Document doc, AceDocument aceDoc) {
ArrayList<AceEntity> entities = aceDoc.entities;
for (int i = 0; i < entities.size(); i++) {
AceEntity entity = (AceEntity) entities.get(i);
ArrayList<AceEntityMention> mentions = entity.mentions;
for (int j = 0; j < mentions.size(); j++) {
AceEntityMention mention = mentions.get(j);
// we compute a jetSpan not including trailing whitespace
Span aceSpan = mention.head;
Span jetSpan = new Span(aceSpan.start(), aceSpan.end() + 1);
FeatureSet features = new FeatureSet("entity", new Integer(i));
if (showTypes) {
features.put("type", entity.type.substring(0, 3));
if (entity.subtype != null) features.put("subtype", entity.subtype);
}
doc.annotate("mention", jetSpan, features);
}
}
}
/** write 'fileText' out as file 'XMLfileName' with ENAMEX tags for the names in the document */
static void addENAMEXtags(Document doc, AceDocument aceDoc) {
ArrayList<AceEntity> entities = aceDoc.entities;
for (int i = 0; i < entities.size(); i++) {
AceEntity entity = entities.get(i);
ArrayList<AceEntityName> names = entity.names;
for (int j = 0; j < names.size(); j++) {
AceEntityName name = names.get(j);
Span aceSpan = name.extent;
Span jetSpan = new Span(aceSpan.start(), aceSpan.end() + 1);
doc.annotate("ENAMEX", jetSpan, new FeatureSet("TYPE", entity.type));
}
// for 2004 we have to examine PRE mentions and decide which are names
if (year.equals("2004")) {
ArrayList<AceEntityMention> mentions = entity.mentions;
for (int j = 0; j < mentions.size(); j++) {
AceEntityMention mention = mentions.get(j);
String htext = Resolve.normalizeName(mention.headText);
String[] mentionName = Gazetteer.splitAtWS(htext);
String preClass = preDict.get(htext.toLowerCase());
if (mention.type.equals("PRE")) {
if (gazetteer.isNationality(mentionName)
|| gazetteer.isLocation(mentionName)
|| "N".equals(preClass)) {
Span aceSpan = mention.head;
Span jetSpan = new Span(aceSpan.start(), aceSpan.end() + 1);
doc.annotate("ENAMEX", jetSpan, new FeatureSet("TYPE", entity.type));
} else if (preClass != null) {
// do nothing
} else {
System.out.println(
"Unclassified PRE: " + mention.text + " {" + mention.headText + ")");
unknownPre.add(htext.toLowerCase());
}
}
}
}
}
}
/** 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);
}
}