Examples with TextAnnotation edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation used on opensource projects

Example 6 with TextAnnotation

use of edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation in project cogcomp-nlp by CogComp.

the class LabelTwoBefore method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    String classifier;
    String prefix = "LabelTwoBefore";
    TextAnnotation ta = c.getTextAnnotation();
    int start = c.getStartSpan() - 2;
    int end = c.getEndSpan() - 2;
    Set<Feature> features = new LinkedHashSet<>();
    for (int i = start; i < end; i++) {
        if (!isPOSFromCounting) {
            classifier = prefix + "_" + "POS";
            if (i >= 0) {
                TokenLabelView POSView = (TokenLabelView) ta.getView(ViewNames.POS);
                String form = ta.getToken(i);
                String tag = POSView.getLabel(i);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            } else
                features.add(new DiscreteFeature(classifier + ":" + ""));
        } else if (isBaseLineCounting) {
            classifier = prefix + "_" + "BaselinePOS";
            if (i >= 0) {
                String form = ta.getToken(i);
                String tag = counter.tag(i, ta);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            } else
                features.add(new DiscreteFeature(classifier + ":" + ""));
        } else {
            classifier = prefix + "_" + "MikheevPOS";
            if (i >= 0) {
                String form = ta.getToken(i);
                String tag = counter.tag(i, ta);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            } else
                features.add(new DiscreteFeature(classifier + ":" + ""));
        }
    }
    return features;
}

Example 7 with TextAnnotation

use of edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation in project cogcomp-nlp by CogComp.

the class POSBaseLineFeatureExtractor method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    String classifier = "BaseLinePOS";
    TextAnnotation ta = c.getTextAnnotation();
    int start = c.getStartSpan();
    int end = c.getEndSpan();
    Set<Feature> features = new LinkedHashSet<>();
    for (int i = start; i < end; i++) {
        String form = ta.getToken(i);
        String tag = counter.tag(i, ta);
        features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
    }
    return features;
}

Example 8 with TextAnnotation

use of edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation in project cogcomp-nlp by CogComp.

the class POSWindowTwo method getFeatures.

@Override
public /**
 * This feature extractor assumes that the TOKEN View, POS View have been
 * generated in the Constituents TextAnnotation. It will use its own POS tag of the
 * two context words before and after the constituent.
 */
Set<Feature> getFeatures(Constituent c) throws EdisonException {
    TextAnnotation ta = c.getTextAnnotation();
    View TOKENS = null, POS = null;
    try {
        TOKENS = ta.getView(ViewNames.TOKENS);
        POS = ta.getView(ViewNames.POS);
    } catch (Exception e) {
        e.printStackTrace();
    }
    // We can assume that the constituent in this case is a Word(Token) described by the LBJ
    // chunk definition
    int startspan = c.getStartSpan();
    int endspan = c.getEndSpan();
    // All our constituents are words(tokens)
    // words two before & after
    int k = 2;
    String[] tags = getwindowtagskfrom(TOKENS, POS, startspan, endspan, k);
    String classifier = "POSWindowTwo";
    String __id, __value;
    Set<Feature> __result = new LinkedHashSet<Feature>();
    for (int i = 0; i < tags.length; i++) {
        if (tags[i] == null) {
            continue;
        } else {
            __id = classifier + ":" + i;
            __value = "(" + tags[i] + ")";
            logger.info(__id + __value);
            __result.add(new DiscreteFeature(__id + __value));
        }
    }
    return __result;
}

Example 9 with TextAnnotation

use of edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation in project cogcomp-nlp by CogComp.

the class PPFeatures method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    TextAnnotation ta = c.getTextAnnotation();
    TreeView parse = (TreeView) ta.getView(parseViewName);
    Set<Feature> feats = new HashSet<>();
    try {
        Constituent phrase = parse.getParsePhrase(c);
        // if the phrase is a PP, then the head word of its
        // rightmost NP child.
        List<Relation> rels = phrase.getOutgoingRelations();
        for (int i = rels.size() - 1; i >= 0; i--) {
            Relation relation = rels.get(i);
            if (relation == null)
                continue;
            Constituent target = relation.getTarget();
            if (ParseTreeProperties.isNominal(target.getLabel())) {
                int head = CollinsHeadFinder.getInstance().getHeadWordPosition(phrase);
                feats.add(DiscreteFeature.create("np-head:" + ta.getToken(head).toLowerCase()));
                feats.add(DiscreteFeature.create("np-head-pos:" + WordHelpers.getPOS(ta, head)));
                break;
            }
        }
        // if the phrase's parent is a PP, then the head of that PP.
        Constituent parent = phrase.getIncomingRelations().get(0).getSource();
        if (parent.getLabel().equals("PP")) {
            int head = CollinsHeadFinder.getInstance().getHeadWordPosition(phrase);
            feats.add(DiscreteFeature.create("p-head:" + ta.getToken(head).toLowerCase()));
        }
    } catch (EdisonException e) {
        throw new RuntimeException(e);
    } catch (Exception e) {
        e.printStackTrace();
    }
    return feats;
}

Example 10 with TextAnnotation

use of edu.illinois.cs.cogcomp.core.datastructures.textannotation.TextAnnotation in project cogcomp-nlp by CogComp.

the class ProjectedPath method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    TextAnnotation ta = c.getTextAnnotation();
    TreeView parse = (TreeView) ta.getView(parseViewName);
    Set<Feature> feats = new HashSet<>();
    // Clone this to avoid concurrency problems
    Constituent c2 = null;
    try {
        c2 = parse.getParsePhrase(c).cloneForNewView("");
    } catch (Exception e) {
        e.printStackTrace();
    }
    assert c2 != null;
    if (!c2.getLabel().equals("VP"))
        return feats;
    boolean found = false;
    boolean done = false;
    while (!done) {
        List<Relation> rels = c2.getIncomingRelations();
        if (rels.size() == 0)
            done = true;
        else {
            Constituent parent = rels.get(0).getSource();
            if (parent.getLabel().equals("VP")) {
                found = true;
                c2 = parent;
            } else {
                done = true;
            }
        }
    }
    if (found) {
        // Clone this to avoid concurrency problems
        Constituent c1 = null;
        try {
            c1 = parse.getParsePhrase(c.getIncomingRelations().get(0).getSource()).cloneForNewView("");
        } catch (Exception e) {
            e.printStackTrace();
        }
        assert c1 != null;
        String path = PathFeatureHelper.getFullParsePathString(c1, c2, 400);
        feats.add(DiscreteFeature.create(path));
    }
    return feats;
}