Examples with Feature org.dkpro.tc.api.features.Feature used on opensource projects

Example 11 with Feature

use of org.dkpro.tc.api.features.Feature in project dkpro-tc by dkpro.

the class SimilarityPairFeatureExtractor method extract.

@Override
public Set<Feature> extract(JCas view1, JCas view2) throws TextClassificationException {
    try {
        double similarity;
        switch(textSimilarityResource.getMode()) {
            case text:
                similarity = textSimilarityResource.getSimilarity(view1.getDocumentText(), view2.getDocumentText());
                break;
            case jcas:
                similarity = ((JCasTextSimilarityMeasure) textSimilarityResource).getSimilarity(view1, view2);
                break;
            default:
                List<String> f1 = getItems(view1);
                List<String> f2 = getItems(view2);
                // Remove "_" tokens
                for (int i = f1.size() - 1; i >= 0; i--) {
                    if (f1.get(i) == null || f1.get(i).equals("_")) {
                        f1.remove(i);
                    }
                }
                for (int i = f2.size() - 1; i >= 0; i--) {
                    if (f2.get(i) == null || f2.get(i).equals("_")) {
                        f2.remove(i);
                    }
                }
                similarity = textSimilarityResource.getSimilarity(f1, f2);
        }
        return new Feature("Similarity" + textSimilarityResource.getName(), similarity, FeatureType.NUMERIC).asSet();
    } catch (FeaturePathException e) {
        throw new TextClassificationException(e);
    } catch (SimilarityException e) {
        throw new TextClassificationException(e);
    }
}

Example 12 with Feature

use of org.dkpro.tc.api.features.Feature in project dkpro-tc by dkpro.

the class SimilarityPairFeatureTest method similarityPairFeatureTest.

@Test
public void similarityPairFeatureTest() throws Exception {
    ExternalResourceDescription gstResource = ExternalResourceFactory.createExternalResourceDescription(GreedyStringTilingMeasureResource.class, GreedyStringTilingMeasureResource.PARAM_MIN_MATCH_LENGTH, "3");
    AnalysisEngineDescription desc = createEngineDescription(NoOpAnnotator.class);
    AnalysisEngine engine = createEngine(desc);
    JCas jcas = engine.newJCas();
    TokenBuilder<Token, Sentence> tb = new TokenBuilder<Token, Sentence>(Token.class, Sentence.class);
    JCas view1 = jcas.createView(VIEW1);
    view1.setDocumentLanguage("en");
    tb.buildTokens(view1, "This is a test .");
    JCas view2 = jcas.createView(VIEW2);
    view2.setDocumentLanguage("en");
    tb.buildTokens(view2, "Test is this .");
    engine.process(jcas);
    SimilarityPairFeatureExtractor extractor = FeatureUtil.createResource(SimilarityPairFeatureExtractor.class, SimilarityPairFeatureExtractor.PARAM_UNIQUE_EXTRACTOR_NAME, "123", SimilarityPairFeatureExtractor.PARAM_SEGMENT_FEATURE_PATH, Token.class.getName(), SimilarityPairFeatureExtractor.PARAM_TEXT_SIMILARITY_RESOURCE, gstResource);
    Set<Feature> features = extractor.extract(jcas.getView(VIEW1), jcas.getView(VIEW2));
    Assert.assertEquals(1, features.size());
    Iterator<Feature> iter = features.iterator();
    assertFeature("SimilarityGreedyStringTiling_3", 0.8125, iter.next(), 0.0001);
}

Example 13 with Feature

use of org.dkpro.tc.api.features.Feature in project dkpro-tc by dkpro.

the class CosineFeatureExtractor method extract.

@Override
public Set<Feature> extract(JCas view1, JCas view2) throws TextClassificationException {
    try {
        TextClassificationTarget aTarget1 = JCasUtil.selectSingle(view1, TextClassificationTarget.class);
        TextClassificationTarget aTarget2 = JCasUtil.selectSingle(view2, TextClassificationTarget.class);
        // Note: getSimilarity(String, String) is *not* a convenience
        // method for getSimilarity(Collection<String>, Collection<String>).
        Set<String> text1 = NGramUtils.getDocumentNgrams(view1, aTarget1, true, false, 1, 1, stopwords, ngramAnnotationType).getKeys();
        Set<String> text2 = NGramUtils.getDocumentNgrams(view2, aTarget2, true, false, 1, 1, stopwords, ngramAnnotationType).getKeys();
        double similarity = measure.getSimilarity(text1, text2);
        // Temporary fix for DKPro Similarity Issue 30
        if (Double.isNaN(similarity)) {
            similarity = 0.0;
        }
        return new Feature("Similarity" + measure.getName(), similarity, FeatureType.NUMERIC).asSet();
    } catch (SimilarityException e) {
        throw new TextClassificationException(e);
    }
}

Example 14 with Feature

use of org.dkpro.tc.api.features.Feature in project dkpro-tc by dkpro.

the class PhoneticNGram method extract.

@Override
public Set<Feature> extract(JCas jcas, TextClassificationTarget aTarget) throws TextClassificationException {
    Set<Feature> features = new HashSet<Feature>();
    FrequencyDistribution<String> documentNgrams = PhoneticNGramMC.getDocumentPhoneticNgrams(jcas, aTarget, ngramMinN, ngramMaxN);
    for (String topNgram : topKSet.getKeys()) {
        if (documentNgrams.getKeys().contains(topNgram)) {
            features.add(new Feature(getFeaturePrefix() + "_" + topNgram, 1, FeatureType.BOOLEAN));
        } else {
            features.add(new Feature(getFeaturePrefix() + "_" + topNgram, 0, true, FeatureType.BOOLEAN));
        }
    }
    return features;
}

Example 15 with Feature

use of org.dkpro.tc.api.features.Feature in project dkpro-tc by dkpro.

the class CosineSimilarityTest method testCosSimWithLemmas.

@Test
public void testCosSimWithLemmas() throws Exception {
    CosineSimilarityTest test = new CosineSimilarityTest();
    test.initialize();
    test.parameters = new Object[] { CosineFeatureExtractor.PARAM_UNIQUE_EXTRACTOR_NAME, "123", CosineFeatureExtractor.PARAM_SOURCE_LOCATION, test.lucenePath.toString(), IdfPairMetaCollector.PARAM_TARGET_LOCATION, test.lucenePath.toString(), CosineFeatureExtractor.PARAM_NGRAM_ANNO_TYPE, "de.tudarmstadt.ukp.dkpro.core.api.segmentation.type.Lemma" };
    test.runPipeline();
    assertTrue(test.featureNames.first().equals("SimilarityCosineSimilarity"));
    assertEquals(test.featureNames.size(), 1);
    for (Feature feat : test.instanceList.get(0).getFeatures()) {
        assertEquals(0.2, (double) feat.getValue(), epsilon);
    }
}