Examples with UnitizingAnnotationStudy org.dkpro.statistics.agreement.unitizing.UnitizingAnnotationStudy used on opensource projects

Example 1 with UnitizingAnnotationStudy

use of org.dkpro.statistics.agreement.unitizing.UnitizingAnnotationStudy in project webanno by webanno.

the class KrippendorffAlphaUnitizingAgreementMeasure method calculatePairAgreement.

public UnitizingAgreementResult calculatePairAgreement(Map<String, List<CAS>> aCasMap) {
    String typeName = getFeature().getLayer().getName();
    // Calculate a character offset continuum over all CASses. We assume here that the documents
    // all have the same size - since the users cannot change the document sizes, this should be
    // an universally true assumption.
    List<CAS> firstUserCasses = aCasMap.values().stream().findFirst().get();
    int docCount = firstUserCasses.size();
    int[] docSizes = new int[docCount];
    Arrays.fill(docSizes, 0);
    for (Entry<String, List<CAS>> set : aCasMap.entrySet()) {
        int i = 0;
        for (CAS cas : set.getValue()) {
            if (cas != null) {
                assert docSizes[i] == 0 || docSizes[i] == cas.getDocumentText().length();
                docSizes[i] = cas.getDocumentText().length();
            }
            i++;
        }
    }
    int continuumSize = Arrays.stream(docSizes).sum();
    // Create a unitizing study for that continuum.
    UnitizingAnnotationStudy study = new UnitizingAnnotationStudy(continuumSize);
    // them to the unitizing study based on character offsets.
    for (Entry<String, List<CAS>> set : aCasMap.entrySet()) {
        int raterIdx = study.addRater(set.getKey());
        int docOffset = 0;
        int i = 0;
        for (CAS cas : set.getValue()) {
            // skip it.
            if (cas != null) {
                Type t = cas.getTypeSystem().getType(typeName);
                Feature f = t.getFeatureByBaseName(getFeature().getName());
                int currentDocOffset = docOffset;
                cas.select(t).map(fs -> (AnnotationFS) fs).forEach(fs -> {
                    study.addUnit(currentDocOffset + fs.getBegin(), fs.getEnd() - fs.getBegin(), raterIdx, FSUtil.getFeature(fs, f, Object.class));
                });
            }
            docOffset += docSizes[i];
            i++;
        }
    }
    UnitizingAgreementResult result = new UnitizingAgreementResult(typeName, getFeature().getName(), study, new ArrayList<>(aCasMap.keySet()), getTraits().isExcludeIncomplete());
    IAgreementMeasure agreement = new KrippendorffAlphaUnitizingAgreement(study);
    if (result.getStudy().getUnitCount() > 0) {
        result.setAgreement(agreement.calculateAgreement());
    } else {
        result.setAgreement(Double.NaN);
    }
    return result;
}