Examples with DiffResult de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult used on opensource projects

Example 11 with DiffResult

use of de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult in project webanno by webanno.

the class CasMergeTest method thatIncompleteAnnotationIsMerged.

/**
 * If one annotator has provided an annotation at a given position and the other annotator did
 * not (i.e. the annotations are incomplete), then this should be detected as a disagreement.
 */
@Test
public void thatIncompleteAnnotationIsMerged() throws Exception {
    CAS user1 = CasFactory.createText("word");
    createTokenAndOptionalPos(user1, 0, 4, "X");
    CAS user2 = CasFactory.createText("word");
    createTokenAndOptionalPos(user2, 0, 4, null);
    Map<String, List<CAS>> casByUser = new LinkedHashMap<>();
    casByUser.put("user1", asList(user1));
    casByUser.put("user2", asList(user2));
    JCas curatorCas = createText(casByUser.values().stream().flatMap(Collection::stream).findFirst().get().getDocumentText());
    DiffResult result = doDiff(diffAdapters, LINK_TARGET_AS_LABEL, casByUser).toResult();
    sut.setMergeIncompleteAnnotations(true);
    sut.reMergeCas(result, document, null, curatorCas.getCas(), getSingleCasByUser(casByUser));
    assertThat(result.getDifferingConfigurationSets()).isEmpty();
    assertThat(result.getIncompleteConfigurationSets().values()).extracting(set -> set.getPosition()).usingFieldByFieldElementComparator().containsExactly(new SpanPosition(null, null, 0, POS.class.getName(), 0, 4, "word", null, null, -1, -1, null, null));
    assertThat(select(curatorCas, POS.class)).hasSize(1);
}

Example 12 with DiffResult

use of de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult in project webanno by webanno.

the class CasMergeTest method thatIncompleteAnnotationIsNotMerged.

/**
 * If one annotator has provided an annotation at a given position and the other annotator did
 * not (i.e. the annotations are incomplete), then this should be detected as a disagreement.
 */
@Test
public void thatIncompleteAnnotationIsNotMerged() throws Exception {
    CAS user1 = CasFactory.createText("word");
    createTokenAndOptionalPos(user1, 0, 4, "X");
    CAS user2 = CasFactory.createText("word");
    createTokenAndOptionalPos(user2, 0, 4, null);
    Map<String, List<CAS>> casByUser = new LinkedHashMap<>();
    casByUser.put("user1", asList(user1));
    casByUser.put("user2", asList(user2));
    JCas curatorCas = createText(// 
    casByUser.values().stream().flatMap(// 
    Collection::stream).findFirst().get().getDocumentText());
    DiffResult result = doDiff(diffAdapters, LINK_TARGET_AS_LABEL, casByUser).toResult();
    sut.reMergeCas(result, document, null, curatorCas.getCas(), getSingleCasByUser(casByUser));
    assertThat(result.getDifferingConfigurationSets()).isEmpty();
    assertThat(result.getIncompleteConfigurationSets().values()).extracting(set -> set.getPosition()).usingFieldByFieldElementComparator().containsExactly(new SpanPosition(null, null, 0, POS.class.getName(), 0, 4, "word", null, null, -1, -1, null, null));
    assertThat(select(curatorCas, POS.class)).isEmpty();
}

Example 13 with DiffResult

use of de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult in project webanno by webanno.

the class CasMergeTest method multiLinkWithRoleTargetDifferenceTest.

@Test
public void multiLinkWithRoleTargetDifferenceTest() throws Exception {
    JCas jcasA = createJCas(createMultiLinkWithRoleTestTypeSystem("f1"));
    makeLinkHostFS(jcasA, 0, 0, makeLinkFS(jcasA, "slot1", 0, 0));
    JCas jcasB = createJCas(createMultiLinkWithRoleTestTypeSystem("f1"));
    makeLinkHostFS(jcasB, 0, 0, makeLinkFS(jcasB, "slot1", 10, 10));
    Map<String, List<CAS>> casByUser = new LinkedHashMap<>();
    casByUser.put("user1", asList(jcasA.getCas()));
    casByUser.put("user2", asList(jcasB.getCas()));
    JCas curatorCas = createJCas(createMultiLinkWithRoleTestTypeSystem("f1"));
    curatorCas.setDocumentText(casByUser.values().stream().flatMap(Collection::stream).findFirst().get().getDocumentText());
    DiffResult result = doDiff(diffAdapters, LINK_TARGET_AS_LABEL, casByUser).toResult();
    // result.print(System.out);
    sut.reMergeCas(result, document, null, curatorCas.getCas(), getSingleCasByUser(casByUser));
    Type hostType = curatorCas.getCas().getTypeSystem().getType(HOST_TYPE);
    FeatureSupport slotSupport = featureSupportRegistry.getFeatureSupport(slotFeature);
    assertThat(select(curatorCas.getCas(), hostType)).hasSize(1);
    assertThat(select(curatorCas.getCas(), hostType).stream().map(host -> (List) slotSupport.getFeatureValue(slotFeature, host))).allMatch(Collection::isEmpty);
}

Example 14 with DiffResult

use of de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult in project webanno by webanno.

the class CasMergeTest method multiLinkMultiHostTest.

@Test
public void multiLinkMultiHostTest() throws Exception {
    // Creating two span stacked annotations. This should cause the data not to be merged.
    JCas jcasA = createJCas(createMultiLinkWithRoleTestTypeSystem("f1"));
    makeLinkHostFS(jcasA, 0, 0, makeLinkFS(jcasA, "slot1", 0, 0));
    makeLinkHostFS(jcasA, 0, 0, makeLinkFS(jcasA, "slot1", 0, 0));
    JCas jcasB = createJCas(createMultiLinkWithRoleTestTypeSystem("f1"));
    makeLinkHostFS(jcasB, 0, 0, makeLinkFS(jcasB, "slot1", 0, 0));
    Map<String, List<CAS>> casByUser = new LinkedHashMap<>();
    casByUser.put("user1", asList(jcasA.getCas()));
    casByUser.put("user2", asList(jcasB.getCas()));
    CAS curatorCas = createJCas(createMultiLinkWithRoleTestTypeSystem("f1")).getCas();
    curatorCas.setDocumentText(casByUser.values().stream().flatMap(Collection::stream).findFirst().get().getDocumentText());
    SpanDiffAdapter adapter = new SpanDiffAdapter(HOST_TYPE);
    adapter.addLinkFeature("links", "role", "target");
    DiffResult result = doDiff(asList(adapter), LINK_TARGET_AS_LABEL, casByUser).toResult();
    // result.print(System.out);
    sut.reMergeCas(result, document, null, curatorCas, getSingleCasByUser(casByUser));
    assertThat(select(curatorCas, getType(curatorCas, HOST_TYPE))).isEmpty();
}

Example 15 with DiffResult

use of de.tudarmstadt.ukp.clarin.webanno.curation.casdiff.CasDiff.DiffResult in project webanno by webanno.

the class CasMerge method reMergeCas.

/**
 * Using {@code DiffResult}, determine the annotations to be deleted from the randomly generated
 * MergeCase. The initial Merge CAs is stored under a name {@code CurationPanel#CURATION_USER}.
 * <p>
 * Any similar annotations stacked in a {@code CasDiff2.Position} will be assumed a difference
 * <p>
 * Any two annotation with different value will be assumed a difference
 * <p>
 * Any non stacked empty/null annotations are assumed agreement
 * <p>
 * Any non stacked annotations with similar values for each of the features are assumed
 * agreement
 * <p>
 * Any two link mode / slotable annotations which agree on the base features are assumed
 * agreement
 *
 * @param aDiff
 *            the {@link DiffResult}
 * @param aCases
 *            a map of {@code CAS}s for each users and the random merge
 */
public void reMergeCas(DiffResult aDiff, SourceDocument aTargetDocument, String aTargetUsername, CAS aTargetCas, Map<String, CAS> aCases) throws AnnotationException, UIMAException {
    silenceEvents = true;
    int updated = 0;
    int created = 0;
    Set<LogMessage> messages = new LinkedHashSet<>();
    // Remove any annotations from the target CAS - keep type system, sentences and tokens
    clearAnnotations(aTargetCas);
    // If there is nothing to merge, bail out
    if (aCases.isEmpty()) {
        return;
    }
    // Set up a cache for resolving type to layer to avoid hammering the DB as we process each
    // position
    Map<String, AnnotationLayer> type2layer = aDiff.getPositions().stream().map(Position::getType).distinct().map(type -> schemaService.findLayer(aTargetDocument.getProject(), type)).collect(toMap(AnnotationLayer::getName, identity()));
    List<String> layerNames = new ArrayList<>(type2layer.keySet());
    // Move token layer to front
    if (layerNames.contains(Token.class.getName())) {
        layerNames.remove(Token.class.getName());
        layerNames.add(0, Token.class.getName());
    }
    // Move sentence layer to front
    if (layerNames.contains(Sentence.class.getName())) {
        layerNames.remove(Sentence.class.getName());
        layerNames.add(0, Sentence.class.getName());
    }
    // and sentences before the others)
    for (String layerName : layerNames) {
        List<SpanPosition> positions = aDiff.getPositions().stream().filter(pos -> layerName.equals(pos.getType())).filter(pos -> pos instanceof SpanPosition).map(pos -> (SpanPosition) pos).filter(pos -> pos.getFeature() == null).collect(Collectors.toList());
        if (positions.isEmpty()) {
            continue;
        }
        LOG.debug("Processing {} span positions on layer {}", positions.size(), layerName);
        // Slots are also excluded for the moment
        for (SpanPosition position : positions) {
            LOG.trace(" |   processing {}", position);
            ConfigurationSet cfgs = aDiff.getConfigurationSet(position);
            if (!shouldMerge(aDiff, cfgs)) {
                continue;
            }
            try {
                Map<String, List<CAS>> casMap = new LinkedHashMap<>();
                aCases.forEach((k, v) -> casMap.put(k, asList(v)));
                AnnotationFS sourceFS = (AnnotationFS) cfgs.getConfigurations().get(0).getRepresentative(casMap);
                CasMergeOperationResult result = mergeSpanAnnotation(aTargetDocument, aTargetUsername, type2layer.get(position.getType()), aTargetCas, sourceFS, false);
                LOG.trace(" `-> merged annotation with agreement");
                switch(result.getState()) {
                    case CREATED:
                        created++;
                        break;
                    case UPDATED:
                        updated++;
                        break;
                }
            } catch (AnnotationException e) {
                LOG.trace(" `-> not merged annotation: {}", e.getMessage());
                messages.add(LogMessage.error(this, "%s", e.getMessage()));
            }
        }
    }
    // After the spans are in place, we can merge the slot features
    for (String layerName : layerNames) {
        List<SpanPosition> positions = aDiff.getPositions().stream().filter(pos -> layerName.equals(pos.getType())).filter(pos -> pos instanceof SpanPosition).map(pos -> (SpanPosition) pos).filter(pos -> pos.getFeature() != null).collect(Collectors.toList());
        if (positions.isEmpty()) {
            continue;
        }
        LOG.debug("Processing {} slot positions on layer [{}]", positions.size(), layerName);
        for (SpanPosition position : positions) {
            LOG.trace(" |   processing {}", position);
            ConfigurationSet cfgs = aDiff.getConfigurationSet(position);
            if (!shouldMerge(aDiff, cfgs)) {
                continue;
            }
            try {
                Map<String, List<CAS>> casMap = new LinkedHashMap<>();
                aCases.forEach((k, v) -> casMap.put(k, asList(v)));
                AnnotationFS sourceFS = (AnnotationFS) cfgs.getConfigurations().get(0).getRepresentative(casMap);
                AID sourceFsAid = cfgs.getConfigurations().get(0).getRepresentativeAID();
                mergeSlotFeature(aTargetDocument, aTargetUsername, type2layer.get(position.getType()), aTargetCas, sourceFS, sourceFsAid.feature, sourceFsAid.index);
                LOG.trace(" `-> merged annotation with agreement");
            } catch (AnnotationException e) {
                LOG.trace(" `-> not merged annotation: {}", e.getMessage());
                messages.add(LogMessage.error(this, "%s", e.getMessage()));
            }
        }
    }
    // Finally, we merge the relations
    for (String layerName : layerNames) {
        List<RelationPosition> positions = aDiff.getPositions().stream().filter(pos -> layerName.equals(pos.getType())).filter(pos -> pos instanceof RelationPosition).map(pos -> (RelationPosition) pos).collect(Collectors.toList());
        if (positions.isEmpty()) {
            continue;
        }
        LOG.debug("Processing {} relation positions on layer [{}]", positions.size(), layerName);
        for (RelationPosition position : positions) {
            LOG.trace(" |   processing {}", position);
            ConfigurationSet cfgs = aDiff.getConfigurationSet(position);
            if (!shouldMerge(aDiff, cfgs)) {
                continue;
            }
            try {
                Map<String, List<CAS>> casMap = new LinkedHashMap<>();
                aCases.forEach((k, v) -> casMap.put(k, asList(v)));
                AnnotationFS sourceFS = (AnnotationFS) cfgs.getConfigurations().get(0).getRepresentative(casMap);
                CasMergeOperationResult result = mergeRelationAnnotation(aTargetDocument, aTargetUsername, type2layer.get(position.getType()), aTargetCas, sourceFS, false);
                LOG.trace(" `-> merged annotation with agreement");
                switch(result.getState()) {
                    case CREATED:
                        created++;
                        break;
                    case UPDATED:
                        updated++;
                        break;
                }
            } catch (AnnotationException e) {
                LOG.trace(" `-> not merged annotation: {}", e.getMessage());
                messages.add(LogMessage.error(this, "%s", e.getMessage()));
            }
        }
    }
    if (eventPublisher != null) {
        eventPublisher.publishEvent(new BulkAnnotationEvent(this, aTargetDocument, aTargetUsername, null));
    }
}