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

Example 21 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature 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 22 with Feature

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

the class MixedChunkWindowTwoBeforePOSWindowThreeBefore method getFeatures.

@Override
public /**
     * This feature extractor assumes that the TOKEN View, POS View and the SHALLOW_PARSE View have been
     * generated in the Constituents TextAnnotation. It will use its own POS tag and well as the POS tag
     * and the SHALLOW_PARSE (Chunk) labels of the previous two tokens and return it as a discrete feature. 
     *
     **/
Set<Feature> getFeatures(Constituent c) throws EdisonException {
    TextAnnotation ta = c.getTextAnnotation();
    try {
        TOKENS = ta.getView(ViewNames.TOKENS);
        POS = ta.getView(ViewNames.POS);
        SHALLOW_PARSE = ta.getView(ViewNames.SHALLOW_PARSE);
    } 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
    int k = -2;
    List<Constituent> wordstwobefore = getwordskfrom(TOKENS, startspan, endspan, k);
    if (wordstwobefore.size() != 2)
        return null;
    String[] tags = new String[3];
    String[] labels = new String[2];
    int i = 0;
    for (Constituent token : wordstwobefore) {
        // Should only be one POS tag for each token
        List<String> POS_tag = POS.getLabelsCoveringSpan(token.getStartSpan(), token.getEndSpan());
        List<String> Chunk_label = SHALLOW_PARSE.getLabelsCoveringSpan(token.getStartSpan(), token.getEndSpan());
        if (POS_tag.size() != 1 || Chunk_label.size() != 1) {
            logger.warn("Error token has more than one POS tag or Chunk Label.");
        }
        labels[i] = Chunk_label.get(0);
        tags[i] = POS_tag.get(0);
        i++;
    }
    tags[i] = POS.getLabelsCoveringSpan(startspan, endspan).get(0);
    Set<Feature> __result = new LinkedHashSet<Feature>();
    String classifier = "MixedChunkWindowTwoBeforePOSWindowThreeBefore";
    String __id = classifier + ":" + "ll";
    String __value = "(" + (labels[0] + "_" + labels[1]) + ")";
    /*
         * BufferedWriter output = null; try { File file = new
         * File("/home/pvijaya2/feat-output.txt");
         * 
         * if(!file.exists()){ file.createNewFile(); }
         * 
         * FileWriter fw = new FileWriter(file,true);
         * 
         * //BufferedWriter writer give better performance BufferedWriter bw = new
         * BufferedWriter(fw);
         */
    logger.info(__id + __value);
    __result.add(new DiscreteFeature(__id + __value));
    __id = classifier + ":" + "lt1";
    __value = "(" + (labels[0] + "_" + tags[1]) + ")";
    logger.info(__id + __value);
    __result.add(new DiscreteFeature(__id + __value));
    __id = classifier + ":" + "lt2";
    __value = "" + (labels[1] + "_" + tags[2]);
    logger.info(__id + __value);
    __result.add(new DiscreteFeature(__id + __value));
    return __result;
}

Example 23 with Feature

use of edu.illinois.cs.cogcomp.edison.features.Feature 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 24 with Feature

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

the class TestWordConjunctionOneTwoThreeGramWindowTwo method test.

@Test
public final void test() throws EdisonException {
    // Using the 3rd constituent as a test
    List<Constituent> testList = ta.getView("TOKENS").getConstituents();
    Constituent test = testList.get(3);
    WordConjunctionOneTwoThreeGramWindowTwo fex = new WordConjunctionOneTwoThreeGramWindowTwo("WordConj3GramWin2");
    Set<Feature> feats = fex.getFeatures(test);
    String[] expected_outputs = { "WordConjunctionOneTwoThreeGramWindowTwo:-2_1(construction)", "WordConjunctionOneTwoThreeGramWindowTwo:-1_1(of)", "WordConjunctionOneTwoThreeGramWindowTwo:0_1(the)", "WordConjunctionOneTwoThreeGramWindowTwo:1_1(John)", "WordConjunctionOneTwoThreeGramWindowTwo:2_1(Smith)", "WordConjunctionOneTwoThreeGramWindowTwo:-2_2(construction_of)", "WordConjunctionOneTwoThreeGramWindowTwo:-1_2(of_the)", "WordConjunctionOneTwoThreeGramWindowTwo:0_2(the_John)", "WordConjunctionOneTwoThreeGramWindowTwo:1_2(John_Smith)", "WordConjunctionOneTwoThreeGramWindowTwo:2_2(Smith)", "WordConjunctionOneTwoThreeGramWindowTwo:-2_3(construction_of_the)", "WordConjunctionOneTwoThreeGramWindowTwo:-1_3(of_the_John)", "WordConjunctionOneTwoThreeGramWindowTwo:0_3(the_John_Smith)", "WordConjunctionOneTwoThreeGramWindowTwo:1_3(John_Smith)", "WordConjunctionOneTwoThreeGramWindowTwo:2_3(Smith)" };
    if (feats == null)
        fail("Feats are returning NULL.");
    for (Feature f : feats) {
        assertTrue(ArrayUtils.contains(expected_outputs, f.getName()));
    }
}

Example 25 with Feature

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

the class TestBrownClusterFeatureExtractor method test.

@Test
public final void test() {
    int[] prefixLengths = new int[] { 4, 6, 10, 20 };
    BrownClusterFeatureExtractor bcfex1 = BrownClusterFeatureExtractor.instance1000;
    BrownClusterFeatureExtractor bcfex2 = null;
    try {
        bcfex2 = new BrownClusterFeatureExtractor("bllip", "brownBllipClusters", prefixLengths);
    } catch (EdisonException e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
    BrownClusterFeatureExtractor bcfex3 = null;
    try {
        bcfex3 = new BrownClusterFeatureExtractor("wiki", "brown-english-wikitext.case-intact.txt-c1000-freq10-v3.txt", prefixLengths);
    } catch (EdisonException e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
    TokenizerTextAnnotationBuilder taBldr = new TokenizerTextAnnotationBuilder(new StatefulTokenizer());
    TextAnnotation ta = taBldr.createTextAnnotation("test", "test", "This test sentence has Joynt and Lieberknecht and Fibonnaci in it " + "just to exercise possible brown cluster hits in resources used by NER.");
    Set<Feature> feats = new HashSet<>();
    for (int wordIndex = 0; wordIndex < ta.size(); ++wordIndex) try {
        feats.addAll(bcfex1.getWordFeatures(ta, wordIndex));
        feats.addAll(bcfex2.getWordFeatures(ta, wordIndex));
        feats.addAll(bcfex3.getWordFeatures(ta, wordIndex));
    } catch (EdisonException e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
    assertTrue(ta.hasView(ViewNames.BROWN_CLUSTERS + "_wiki"));
    String[] featArray = new String[feats.size()];
    int i = 0;
    for (Feature f : feats) featArray[i++] = f.toString();
    Arrays.sort(featArray);
    String actualOutput = StringUtils.join(",", featArray);
    assertEquals(expectedOutput, actualOutput);
}