Examples with Feature edu.illinois.cs.cogcomp.edison.features.Feature used on opensource projects

Example 51 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature in project cogcomp-nlp by CogComp.

the class POSMikheevFeatureExtractor method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    String classifier = "MikheevPOS";
    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 52 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature in project cogcomp-nlp by CogComp.

the class POSWindow method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    String classifier;
    String prefix = "POSWindow";
    TextAnnotation ta = c.getTextAnnotation();
    int lenOfTokens = ta.getTokens().length;
    int start = c.getStartSpan();
    int end = c.getEndSpan();
    Set<Feature> features = new LinkedHashSet<>();
    for (int i = start; i < end; i++) {
        int curr = i, last = i;
        // for (; last <= i+2 && last < lenOfTokens; ++last)
        for (int j = 0; j < 2 && curr > 0; j++) curr -= 1;
        for (int j = 0; j < 2 && last < lenOfTokens - 1; j++) last += 1;
        if (!isPOSFromCounting) {
            classifier = prefix + "_" + "POS";
            for (; curr <= last; curr++) {
                TokenLabelView POSView = (TokenLabelView) ta.getView(ViewNames.POS);
                String form = ta.getToken(curr);
                String tag = POSView.getLabel(curr);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            }
        } else if (isBaseLineCounting) {
            classifier = prefix + "_" + "BaselinePOS";
            for (; curr <= last; curr++) {
                String form = ta.getToken(curr);
                String tag = counter.tag(curr, ta);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            }
        } else {
            classifier = prefix + "_" + "MikheevPOS";
            for (; curr <= last; curr++) {
                String form = ta.getToken(curr);
                String tag = counter.tag(curr, ta);
                features.add(new DiscreteFeature(classifier + ":" + tag + "_" + form));
            }
        }
    }
    return features;
}

Example 53 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature in project cogcomp-nlp by CogComp.

the class PosWordConjunctionSizeTwoWindowSizeTwo 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 and well 
     * as the form of the word as a forms of the words around the constitent a 
     *
     **/
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;
    int window = 2;
    String[] forms = getWindowK(TOKENS, startspan, endspan, k);
    String[] tags = getWindowKTags(POS, startspan, endspan, k);
    String classifier = "PosWordConjunctionSizeTwoWindowSizeTwo";
    String id, value;
    Set<Feature> result = new LinkedHashSet<>();
    for (int j = 0; j < k; j++) {
        for (int i = 0; i < tags.length; i++) {
            StringBuilder f = new StringBuilder();
            for (int context = 0; context <= j && i + context < tags.length; context++) {
                if (context != 0) {
                    f.append("_");
                }
                f.append(tags[i + context]);
                f.append("-");
                f.append(forms[i + context]);
            }
            // 2 is the center object in the array so i should go from -2 to +2 (with 0 being
            // the center)
            // j is the size of the n-gram so it goes 1 to 2
            id = classifier + ":" + ((i - window) + "_" + (j + 1));
            value = "(" + (f.toString()) + ")";
            result.add(new DiscreteFeature(id + value));
        }
    }
    return result;
}

Example 54 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature in project cogcomp-nlp by CogComp.

the class HyphenTagFeature method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    Set<Feature> features = new HashSet<>();
    String surfaceString = c.getSurfaceForm();
    if (surfaceString.contains("-") && c.length() == 1) {
        Constituent predicate = c.getIncomingRelations().get(0).getSource();
        String lemma = predicate.getAttribute(PredicateArgumentView.LemmaIdentifier);
        assert lemma != null;
        if (predicate.getSpan().equals(c.getSpan())) {
            features.add(DiscreteFeature.create("pred-token"));
        }
        String[] parts = surfaceString.split("-");
        for (int i = 0; i < parts.length; i++) {
            String part = parts[i];
            if (part.contains(lemma)) {
                features.add(DiscreteFeature.create(i + ":pred"));
            } else {
                String lowerCase = part.toLowerCase();
                features.add(DiscreteFeature.create(lowerCase));
                features.add(DiscreteFeature.create(i + ":" + lowerCase));
            }
        }
    }
    return features;
}

Example 55 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature in project cogcomp-nlp by CogComp.

the class LabelOneAfter method getFeatures.

@Override
public Set<Feature> getFeatures(Constituent c) throws EdisonException {
    String classifier;
    String prefix = "LabelOneAfter";
    TextAnnotation ta = c.getTextAnnotation();
    int lenOfTokens = ta.getTokens().length;
    int start = c.getStartSpan() + 1;
    int end = c.getEndSpan() + 1;
    Set<Feature> features = new LinkedHashSet<>();
    for (int i = start; i < end; i++) {
        if (!isPOSFromCounting) {
            classifier = prefix + "_" + "POS";
            if (i < lenOfTokens) {
                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 < lenOfTokens) {
                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 < lenOfTokens) {
                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;
}