use of edu.illinois.cs.cogcomp.edison.utilities.EdisonException in project cogcomp-nlp by CogComp.
the class GetParseRightSibling method transform.
@Override
public List<Constituent> transform(Constituent input) {
TextAnnotation ta = input.getTextAnnotation();
TreeView parse = (TreeView) ta.getView(parseViewName);
List<Constituent> siblings = new ArrayList<>();
try {
Constituent phrase = parse.getParsePhrase(input);
List<Relation> in = phrase.getIncomingRelations();
if (in.size() > 0) {
List<Relation> outgoingRelations = in.get(0).getSource().getOutgoingRelations();
int id = -1;
for (int i = 0; i < outgoingRelations.size(); i++) {
Relation r = outgoingRelations.get(i);
if (r.getTarget() == phrase) {
id = i;
break;
}
}
if (id >= 0 && id + 1 < outgoingRelations.size())
siblings.add(outgoingRelations.get(id + 1).getTarget());
}
} catch (EdisonException e) {
throw new RuntimeException(e);
} catch (Exception e) {
e.printStackTrace();
}
return siblings;
}
use of edu.illinois.cs.cogcomp.edison.utilities.EdisonException in project cogcomp-nlp by CogComp.
the class ParsePath method getFeatures.
@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
TextAnnotation ta = c.getTextAnnotation();
TreeView parse = (TreeView) ta.getView(parseViewName);
Set<Feature> features = new LinkedHashSet<>();
List<Relation> incomingRelations = c.getIncomingRelations();
if (incomingRelations.size() > 0) {
Constituent c1, c2;
try {
c1 = parse.getParsePhrase(incomingRelations.get(0).getSource());
c2 = parse.getParsePhrase(c);
} catch (Exception e) {
throw new EdisonException(e);
}
Pair<List<Constituent>, List<Constituent>> paths = PathFeatureHelper.getPathsToCommonAncestor(c1, c2, 400);
List<Constituent> list = new ArrayList<>();
for (int i = 0; i < paths.getFirst().size() - 1; i++) {
list.add(paths.getFirst().get(i));
}
Constituent top = paths.getFirst().get(paths.getFirst().size() - 1);
list.add(top);
for (int i = paths.getSecond().size() - 2; i >= 0; i--) {
list.add(paths.getSecond().get(i));
}
StringBuilder sb = new StringBuilder();
for (int i = 0; i < paths.getFirst().size() - 1; i++) {
Constituent cc = paths.getFirst().get(i);
sb.append(cc.getLabel());
sb.append(PathFeatureHelper.PATH_UP_STRING);
}
String pathToAncestor = sb.toString();
String pathString = PathFeatureHelper.getPathString(paths, true, false);
features.add(DiscreteFeature.create(pathString));
features.add(DiscreteFeature.create(pathToAncestor));
features.add(RealFeature.create("l", list.size()));
}
return features;
}
use of edu.illinois.cs.cogcomp.edison.utilities.EdisonException in project cogcomp-nlp by CogComp.
the class ParsePhraseType method getFeatures.
@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
TextAnnotation ta = c.getTextAnnotation();
TreeView tree = (TreeView) ta.getView(parseViewname);
Constituent phrase;
try {
phrase = tree.getParsePhrase(c);
} catch (Exception e) {
throw new EdisonException(e);
}
Set<Feature> features = new LinkedHashSet<>();
if (phrase != null) {
features.add(DiscreteFeature.create(phrase.getLabel()));
String parentLabel = "ROOT";
if (phrase.getIncomingRelations().size() > 0) {
Constituent parent = phrase.getIncomingRelations().get(0).getSource();
parentLabel = parent.getLabel();
int parentHead = CollinsHeadFinder.getInstance().getHeadWordPosition(parent);
features.add(DiscreteFeature.create("pt:h:" + ta.getToken(parentHead).toLowerCase().trim()));
features.add(DiscreteFeature.create("pt:h-pos:" + WordHelpers.getPOS(ta, parentHead)));
}
features.add(DiscreteFeature.create("pt:" + parentLabel));
}
return features;
}
use of edu.illinois.cs.cogcomp.edison.utilities.EdisonException in project cogcomp-nlp by CogComp.
the class ParseSiblings method getFeatures.
@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
TextAnnotation ta = c.getTextAnnotation();
TreeView parse = (TreeView) ta.getView(parseViewName);
Constituent phrase;
try {
phrase = parse.getParsePhrase(c);
} catch (Exception e) {
throw new EdisonException(e);
}
Set<Feature> features = new LinkedHashSet<>();
if (phrase.getIncomingRelations().size() == 0) {
features.add(DiscreteFeature.create("ONLY_CHILD"));
} else {
Relation incomingEdge = phrase.getIncomingRelations().get(0);
Constituent parent = incomingEdge.getSource();
int position = -1;
for (int i = 0; i < parent.getOutgoingRelations().size(); i++) {
if (parent.getOutgoingRelations().get(i) == incomingEdge) {
position = i;
break;
}
}
assert position >= 0;
if (position == 0)
features.add(DiscreteFeature.create("FIRST_CHILD"));
else if (position == parent.getOutgoingRelations().size() - 1)
features.add(DiscreteFeature.create("LAST_CHILD"));
if (position != 0) {
Constituent sibling = parent.getOutgoingRelations().get(position - 1).getTarget();
String phraseType = sibling.getLabel();
int headWord = CollinsHeadFinder.getInstance().getHeadWordPosition(sibling);
String token = ta.getToken(headWord).toLowerCase().trim();
String pos = WordHelpers.getPOS(ta, headWord);
features.add(DiscreteFeature.create("lsis.pt:" + phraseType));
features.add(DiscreteFeature.create("lsis.hw:" + token));
features.add(DiscreteFeature.create("lsis.hw.pos:" + pos));
}
if (position != parent.getOutgoingRelations().size() - 1) {
Constituent sibling = parent.getOutgoingRelations().get(position + 1).getTarget();
String phraseType = sibling.getLabel();
int headWord = CollinsHeadFinder.getInstance().getHeadWordPosition(sibling);
String token = ta.getToken(headWord).toLowerCase().trim();
String pos = WordHelpers.getPOS(ta, headWord);
features.add(DiscreteFeature.create("rsis.pt:" + phraseType));
features.add(DiscreteFeature.create("rsis.hw:" + token));
features.add(DiscreteFeature.create("rsis.hw.pos:" + pos));
}
}
return features;
}
use of edu.illinois.cs.cogcomp.edison.utilities.EdisonException in project cogcomp-nlp by CogComp.
the class RogetThesaurusFeatures method getFeatures.
@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
if (!loaded) {
try {
// not load the data from classpath; instead using the datastore
// loadFromClassPath();
loadFromDatastore();
} catch (Exception e) {
throw new EdisonException(e);
}
}
String s = c.getTokenizedSurfaceForm().trim();
Set<Feature> features = new LinkedHashSet<>();
if (map.containsKey(s)) {
for (int i : map.get(s)) {
features.add(DiscreteFeature.create(this.id2ClassName.get(i)));
}
} else if (map.containsKey(s.toLowerCase())) {
for (int i : map.get(s.toLowerCase())) {
features.add(DiscreteFeature.create(this.id2ClassName.get(i)));
}
}
return features;
}
Aggregations