package org.apache.ctakes.relationextractor.ae.features;
import java.util.ArrayList;
import java.util.List;
import org.apache.ctakes.constituency.parser.treekernel.TreeExtractor;
import org.apache.ctakes.constituency.parser.util.AnnotationTreeUtils;
import org.apache.ctakes.typesystem.type.syntax.BaseToken;
import org.apache.ctakes.typesystem.type.syntax.TreebankNode;
import org.apache.ctakes.typesystem.type.textsem.IdentifiedAnnotation;
import org.apache.ctakes.utils.tree.SimpleTree;
import org.apache.uima.analysis_engine.AnalysisEngineProcessException;
import org.apache.uima.jcas.JCas;
import org.cleartk.classifier.Feature;
import org.uimafit.util.JCasUtil;
public class FlatTreeFeatureExtractor implements RelationFeaturesExtractor {
// Returns flat POS-trees a la Hovy et al 2012 (EACL)
@Override
public List<Feature> extract(JCas jcas, IdentifiedAnnotation arg1,
IdentifiedAnnotation arg2) throws AnalysisEngineProcessException {
List<Feature> features = new ArrayList<Feature>();
SimpleTree tree = null;
tree = new SimpleTree("BOP");
TreebankNode lca = TreeExtractor.getLCA(AnnotationTreeUtils.annotationNode(jcas, arg1),
AnnotationTreeUtils.annotationNode(jcas, arg2));
SimpleTree arg1Tree = new SimpleTree("ARG1");
SimpleTree arg2Tree = new SimpleTree("ARG2");
tree.addChild(arg1Tree);
List<BaseToken> coveredTokens = JCasUtil.selectCovered(jcas, BaseToken.class, lca);
for(BaseToken token : coveredTokens){
// create pre-terminal tree
SimpleTree tokenTree = new SimpleTree("TOK");
// tokenTree.addChild(new SimpleTree(token.getCoveredText()));
tokenTree.addChild(new SimpleTree(token.getPartOfSpeech()));
// do we add this to one of the arg trees or to the root?
if(token.getEnd() <= arg1.getEnd()){
arg1Tree.addChild(tokenTree);
}else if(token.getBegin() >= arg2.getBegin()){
arg2Tree.addChild(tokenTree);
}else{
SimpleTree termTree = new SimpleTree("TERM");
termTree.addChild(tokenTree);
tree.addChild(termTree);
}
}
tree.addChild(arg2Tree);
features.add(new Feature("TK_BOP", tree.toString()));
return features;
}
}