protected void processRoot(
SentenceParser.Node root, List<String> sentence, List<String> tags, List<String> target) {
if (root != null) {
TreeElement[] elements = root.getElements();
for (int i = 0; i < elements.length; i++) {
if (elements[i].isLeaf()) {
processLeaf((SentenceParser.Leaf) elements[i], false, OTHER, sentence, tags, target);
} else {
processNode((SentenceParser.Node) elements[i], sentence, tags, target, null);
}
}
}
}
private void processNode(
SentenceParser.Node node,
List<String> sentence,
List<String> tags,
List<String> target,
String inheritedTag) {
String phraseTag = getChunkTag(node);
boolean inherited = false;
if (phraseTag.equals(OTHER) && inheritedTag != null) {
phraseTag = inheritedTag;
inherited = true;
}
TreeElement[] elements = node.getElements();
for (int i = 0; i < elements.length; i++) {
if (elements[i].isLeaf()) {
boolean isIntermediate = false;
String tag = phraseTag;
SentenceParser.Leaf leaf = (SentenceParser.Leaf) elements[i];
String localChunk = getChunkTag(leaf);
if (localChunk != null && !tag.equals(localChunk)) {
tag = localChunk;
}
if (isIntermediate(tags, target, tag) && (inherited || i > 0)) {
isIntermediate = true;
}
if (!isIncludePunctuations()
&& leaf.getFunctionalTag() == null
&& (!(i + 1 < elements.length && elements[i + 1].isLeaf())
|| !(i > 0 && elements[i - 1].isLeaf()))) {
isIntermediate = false;
tag = OTHER;
}
processLeaf(leaf, isIntermediate, tag, sentence, tags, target);
} else {
int before = target.size();
processNode((SentenceParser.Node) elements[i], sentence, tags, target, phraseTag);
// if the child node was of a different type we should break the chunk sequence
for (int j = target.size() - 1; j >= before; j--) {
if (!target.get(j).endsWith("-" + phraseTag)) {
phraseTag = OTHER;
break;
}
}
}
}
}
protected String getChunkTag(SentenceParser.Node node) {
String tag = node.getSyntacticTag();
String phraseTag = tag.substring(tag.lastIndexOf(":") + 1);
while (phraseTag.endsWith("-")) {
phraseTag = phraseTag.substring(0, phraseTag.length() - 1);
}
// maybe we should use only np, vp and pp, but will keep ap and advp.
if (phraseTag.equals("np")
|| phraseTag.equals("vp")
|| phraseTag.equals("pp")
|| phraseTag.equals("ap")
|| phraseTag.equals("advp")
|| phraseTag.equals("adjp")) {
phraseTag = phraseTag.toUpperCase();
} else {
phraseTag = OTHER;
}
return phraseTag;
}
