Merge pull request 'subjects cleaning' (#239) from clean_subjects into beta

Reviewed-on: D-Net/dnet-hadoop#239

dhp-common/src/main/java/eu/dnetlib/dhp/common/vocabulary/Vocabulary.java

@@ -83,4 +83,10 @@ public class Vocabulary implements Serializable {
 			.orElse(null);
 	}
 
+	public Qualifier lookup(String id) {
+		return Optional
+			.ofNullable(getSynonymAsQualifier(id))
+			.orElse(getTermAsQualifier(id));
+	}
+
 }

dhp-common/src/main/java/eu/dnetlib/dhp/common/vocabulary/VocabularyGroup.java

@@ -81,6 +81,13 @@ public class VocabularyGroup implements Serializable {
 		vocs.put(id.toLowerCase(), new Vocabulary(id, name));
 	}
 
+	public Optional<Vocabulary> find(final String vocId) {
+		return Optional
+			.ofNullable(vocId)
+			.map(String::toLowerCase)
+			.map(vocs::get);
+	}
+
 	public void addTerm(final String vocId, final String id, final String name) {
 		if (vocabularyExists(vocId)) {
 			vocs.get(vocId.toLowerCase()).addTerm(id, name);

dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/GraphCleaningFunctions.java

@@ -1,6 +1,8 @@
 
 package eu.dnetlib.dhp.schema.oaf.utils;
 
+import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.getProvenance;
+
 import java.time.LocalDate;
 import java.time.ZoneId;
 import java.time.format.DateTimeFormatter;
@@ -16,7 +18,6 @@ import com.github.sisyphsu.dateparser.DateParserUtils;
 import com.google.common.collect.Lists;
 import com.google.common.collect.Sets;
 
-import eu.dnetlib.dhp.common.vocabulary.Vocabulary;
 import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.common.ModelSupport;
@@ -191,8 +192,8 @@ public class GraphCleaningFunctions extends CleaningFunctions {
 						qualifier("und", "Undetermined", ModelConstants.DNET_LANGUAGES));
 			}
 			if (Objects.nonNull(r.getSubject())) {
-				r
+				List<Subject> subjects = Lists
-					.setSubject(
+					.newArrayList(
 						r
 							.getSubject()
 							.stream()
@@ -201,7 +202,18 @@ public class GraphCleaningFunctions extends CleaningFunctions {
 							.filter(sp -> Objects.nonNull(sp.getQualifier()))
 							.filter(sp -> StringUtils.isNotBlank(sp.getQualifier().getClassid()))
 							.map(GraphCleaningFunctions::cleanValue)
-							.collect(Collectors.toList()));
+							.collect(
+								Collectors
+									.toMap(
+										s -> Optional
+											.ofNullable(s.getQualifier())
+											.map(q -> q.getClassid() + s.getValue())
+											.orElse(s.getValue()),
+										Function.identity(),
+										(s1, s2) -> Collections
+											.min(Lists.newArrayList(s1, s1), new SubjectProvenanceComparator())))
+							.values());
+				r.setSubject(subjects);
 			}
 			if (Objects.nonNull(r.getTitle())) {
 				r
@@ -382,14 +394,7 @@ public class GraphCleaningFunctions extends CleaningFunctions {
 									.filter(p -> StringUtils.isNotBlank(p.getValue()))
 									.map(p -> {
 										// hack to distinguish orcid from orcid_pending
-										String pidProvenance = Optional
+										String pidProvenance = getProvenance(p.getDataInfo());
-											.ofNullable(p.getDataInfo())
-											.map(
-												d -> Optional
-													.ofNullable(d.getProvenanceaction())
-													.map(Qualifier::getClassid)
-													.orElse(""))
-											.orElse("");
 										if (p
 											.getQualifier()
 											.getClassid()
@@ -520,6 +525,11 @@ public class GraphCleaningFunctions extends CleaningFunctions {
 		return s;
 	}
 
+	protected static Subject cleanValue(Subject s) {
+		s.setValue(s.getValue().replaceAll(CLEANING_REGEX, " "));
+		return s;
+	}
+
 	protected static Field<String> cleanValue(Field<String> s) {
 		s.setValue(s.getValue().replaceAll(CLEANING_REGEX, " "));
 		return s;

dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/OafMapperUtils.java

@@ -14,6 +14,7 @@ import java.util.stream.Collectors;
 import org.apache.commons.lang3.StringUtils;
 
 import eu.dnetlib.dhp.schema.common.AccessRightComparator;
+import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.common.ModelSupport;
 import eu.dnetlib.dhp.schema.oaf.*;
 
@@ -141,7 +142,7 @@ public class OafMapperUtils {
 	}
 
 	public static Qualifier unknown(final String schemeid, final String schemename) {
-		return qualifier("UNKNOWN", "Unknown", schemeid, schemename);
+		return qualifier(UNKNOWN, "Unknown", schemeid, schemename);
 	}
 
 	public static AccessRight accessRight(
@@ -189,6 +190,17 @@ public class OafMapperUtils {
 		return q;
 	}
 
+	public static Subject subject(
+		final String value,
+		final String classid,
+		final String classname,
+		final String schemeid,
+		final String schemename,
+		final DataInfo dataInfo) {
+
+		return subject(value, qualifier(classid, classname, schemeid, schemename), dataInfo);
+	}
+
 	public static StructuredProperty structuredProperty(
 		final String value,
 		final String classid,
@@ -200,6 +212,20 @@ public class OafMapperUtils {
 		return structuredProperty(value, qualifier(classid, classname, schemeid, schemename), dataInfo);
 	}
 
+	public static Subject subject(
+		final String value,
+		final Qualifier qualifier,
+		final DataInfo dataInfo) {
+		if (value == null) {
+			return null;
+		}
+		final Subject s = new Subject();
+		s.setValue(value);
+		s.setQualifier(qualifier);
+		s.setDataInfo(dataInfo);
+		return s;
+	}
+
 	public static StructuredProperty structuredProperty(
 		final String value,
 		final Qualifier qualifier,
@@ -477,4 +503,15 @@ public class OafMapperUtils {
 		rel.setProperties(properties);
 		return rel;
 	}
+
+	public static String getProvenance(DataInfo dataInfo) {
+		return Optional
+			.ofNullable(dataInfo)
+			.map(
+				d -> Optional
+					.ofNullable(d.getProvenanceaction())
+					.map(Qualifier::getClassid)
+					.orElse(""))
+			.orElse("");
+	}
 }

dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/SubjectProvenanceComparator.java

@@ -0,0 +1,46 @@
+
+package eu.dnetlib.dhp.schema.oaf.utils;
+
+import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.getProvenance;
+import static org.apache.commons.lang3.StringUtils.isBlank;
+
+import java.util.Comparator;
+
+import eu.dnetlib.dhp.schema.oaf.Subject;
+
+public class SubjectProvenanceComparator implements Comparator<Subject> {
+
+	@Override
+	public int compare(Subject left, Subject right) {
+
+		String lProv = getProvenance(left.getDataInfo());
+		String rProv = getProvenance(right.getDataInfo());
+
+		if (isBlank(lProv) && isBlank(rProv))
+			return 0;
+		if (isBlank(lProv))
+			return 1;
+		if (isBlank(rProv))
+			return -1;
+		if (lProv.equals(rProv))
+			return 0;
+		if (lProv.toLowerCase().contains("crosswalk"))
+			return -1;
+		if (rProv.toLowerCase().contains("crosswalk"))
+			return 1;
+		if (lProv.toLowerCase().contains("user"))
+			return -1;
+		if (rProv.toLowerCase().contains("user"))
+			return 1;
+		if (lProv.toLowerCase().contains("propagation"))
+			return -1;
+		if (rProv.toLowerCase().contains("propagation"))
+			return 1;
+		if (lProv.toLowerCase().contains("iis"))
+			return -1;
+		if (rProv.toLowerCase().contains("iis"))
+			return 1;
+
+		return 0;
+	}
+}

dhp-workflows/dhp-aggregation/src/main/java/eu/dnetlib/dhp/actionmanager/Constants.java

@@ -13,6 +13,7 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 import eu.dnetlib.dhp.application.ArgumentApplicationParser;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
+import eu.dnetlib.dhp.schema.oaf.Subject;
 import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
 
 public class Constants {
@@ -58,13 +59,13 @@ public class Constants {
 			.map((MapFunction<String, R>) value -> OBJECT_MAPPER.readValue(value, clazz), Encoders.bean(clazz));
 	}
 
-	public static StructuredProperty getSubject(String sbj, String classid, String classname,
+	public static Subject getSubject(String sbj, String classid, String classname,
 		String diqualifierclassid) {
 		if (sbj.equals(NULL))
 			return null;
-		StructuredProperty sp = new StructuredProperty();
+		Subject s = new Subject();
-		sp.setValue(sbj);
+		s.setValue(sbj);
-		sp
+		s
 			.setQualifier(
 				OafMapperUtils
 					.qualifier(
@@ -72,7 +73,7 @@ public class Constants {
 						classname,
 						ModelConstants.DNET_SUBJECT_TYPOLOGIES,
 						ModelConstants.DNET_SUBJECT_TYPOLOGIES));
-		sp
+		s
 			.setDataInfo(
 				OafMapperUtils
 					.dataInfo(
@@ -88,7 +89,7 @@ public class Constants {
 								ModelConstants.DNET_PROVENANCE_ACTIONS),
 						""));
 
-		return sp;
+		return s;
 
 	}
 }

dhp-workflows/dhp-aggregation/src/main/java/eu/dnetlib/dhp/actionmanager/createunresolvedentities/PrepareFOSSparkJob.java

@@ -24,6 +24,7 @@ import eu.dnetlib.dhp.application.ArgumentApplicationParser;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.oaf.Result;
 import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
+import eu.dnetlib.dhp.schema.oaf.Subject;
 import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
 import eu.dnetlib.dhp.utils.DHPUtils;
 
@@ -79,7 +80,7 @@ public class PrepareFOSSparkJob implements Serializable {
 				HashSet<String> level3 = new HashSet<>();
 				addLevels(level1, level2, level3, first);
 				it.forEachRemaining(v -> addLevels(level1, level2, level3, v));
-				List<StructuredProperty> sbjs = new ArrayList<>();
+				List<Subject> sbjs = new ArrayList<>();
 				level1.forEach(l -> sbjs.add(getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
 				level2.forEach(l -> sbjs.add(getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
 				level3.forEach(l -> sbjs.add(getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));

dhp-workflows/dhp-aggregation/src/main/java/eu/dnetlib/dhp/actionmanager/createunresolvedentities/PrepareSDGSparkJob.java

@@ -24,6 +24,7 @@ import eu.dnetlib.dhp.application.ArgumentApplicationParser;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.oaf.Result;
 import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
+import eu.dnetlib.dhp.schema.oaf.Subject;
 import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
 import eu.dnetlib.dhp.utils.DHPUtils;
 
@@ -73,7 +74,7 @@ public class PrepareSDGSparkJob implements Serializable {
 				Result r = new Result();
 				r.setId(DHPUtils.generateUnresolvedIdentifier(k, DOI));
 				SDGDataModel first = it.next();
-				List<StructuredProperty> sbjs = new ArrayList<>();
+				List<Subject> sbjs = new ArrayList<>();
 				sbjs.add(getSubject(first.getSbj(), SDG_CLASS_ID, SDG_CLASS_NAME, UPDATE_SUBJECT_SDG_CLASS_ID));
 				it
 					.forEachRemaining(

dhp-workflows/dhp-aggregation/src/main/scala/eu/dnetlib/dhp/datacite/DataciteToOAFTransformation.scala

@@ -19,7 +19,7 @@ import java.time.chrono.ThaiBuddhistDate
 import java.time.format.DateTimeFormatter
 import java.util.{Date, Locale}
 import scala.collection.JavaConverters._
-import scala.io.{Codec, Source}
+import scala.io.Source
 
 object DataciteToOAFTransformation {
 
@@ -252,7 +252,7 @@ object DataciteToOAFTransformation {
         .exists(i => i.getHostedby != null && "figshare".equalsIgnoreCase(i.getHostedby.getValue))
       if (hosted_by_figshare) {
         r.getInstance().asScala.foreach(i => i.setAccessright(ModelConstants.OPEN_ACCESS_RIGHT()))
-        val l: List[StructuredProperty] = List()
+        val l: List[Subject] = List()
         r.setSubject(l.asJava)
       }
     }
@@ -492,7 +492,7 @@ object DataciteToOAFTransformation {
       subjects
         .filter(s => s.subject.nonEmpty)
         .map(s =>
-          OafMapperUtils.structuredProperty(
+          OafMapperUtils.subject(
             s.subject.get,
             SUBJ_CLASS,
             SUBJ_CLASS,

dhp-workflows/dhp-aggregation/src/main/scala/eu/dnetlib/dhp/sx/bio/BioDBToOAF.scala

@@ -281,7 +281,7 @@ object BioDBToOAF {
       d.setSubject(
         subjects
           .map(s =>
-            OafMapperUtils.structuredProperty(
+            OafMapperUtils.subject(
               s,
               SUBJ_CLASS,
               SUBJ_CLASS,

dhp-workflows/dhp-aggregation/src/main/scala/eu/dnetlib/dhp/sx/bio/pubmed/PubMedToOaf.scala

@@ -265,8 +265,8 @@ object PubMedToOaf {
         result.setLanguage(term)
     }
 
-    val subjects: List[StructuredProperty] = article.getSubjects.asScala.map(s =>
+    val subjects: List[Subject] = article.getSubjects.asScala.map(s =>
-      OafMapperUtils.structuredProperty(
+      OafMapperUtils.subject(
         s.getValue,
         SUBJ_CLASS,
         SUBJ_CLASS,

dhp-workflows/dhp-aggregation/src/test/java/eu/dnetlib/dhp/actionmanager/createunresolvedentities/ProduceTest.java

@@ -72,7 +72,7 @@ public class ProduceTest {
 
 		JavaRDD<Result> tmp = getResultJavaRDD();
 
-		List<StructuredProperty> sbjs = tmp
+		List<Subject> sbjs = tmp
 			.filter(row -> row.getSubject() != null && row.getSubject().size() > 0)
 			.flatMap(row -> row.getSubject().iterator())
 			.collect();
@@ -169,7 +169,7 @@ public class ProduceTest {
 					.getSubject()
 					.size());
 
-		List<StructuredProperty> sbjs = tmp
+		List<Subject> sbjs = tmp
 			.filter(row -> row.getId().equals(doi))
 			.flatMap(row -> row.getSubject().iterator())
 			.collect();
@@ -396,7 +396,7 @@ public class ProduceTest {
 					.getSubject()
 					.size());
 
-		List<StructuredProperty> sbjs = tmp
+		List<Subject> sbjs = tmp
 			.filter(row -> row.getId().equals(doi))
 			.flatMap(row -> row.getSubject().iterator())
 			.collect();
@@ -508,7 +508,7 @@ public class ProduceTest {
 					.getSubject()
 					.size());
 
-		List<StructuredProperty> sbjs = tmp
+		List<Subject> sbjs = tmp
 			.filter(row -> row.getId().equals(doi))
 			.flatMap(row -> row.getSubject().iterator())
 			.collect();
@@ -537,7 +537,7 @@ public class ProduceTest {
 
 		JavaRDD<Result> tmp = getResultJavaRDDPlusSDG();
 
-		List<StructuredProperty> sbjs_sdg = tmp
+		List<Subject> sbjs_sdg = tmp
 			.filter(row -> row.getSubject() != null && row.getSubject().size() > 0)
 			.flatMap(row -> row.getSubject().iterator())
 			.filter(sbj -> sbj.getQualifier().getClassid().equals(Constants.SDG_CLASS_ID))

dhp-workflows/dhp-broker-events/src/main/java/eu/dnetlib/dhp/broker/oa/util/ConversionUtils.java

@@ -26,20 +26,7 @@ import eu.dnetlib.broker.objects.OaBrokerRelatedPublication;
 import eu.dnetlib.broker.objects.OaBrokerRelatedSoftware;
 import eu.dnetlib.broker.objects.OaBrokerTypedValue;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
-import eu.dnetlib.dhp.schema.oaf.Author;
+import eu.dnetlib.dhp.schema.oaf.*;
-import eu.dnetlib.dhp.schema.oaf.Dataset;
-import eu.dnetlib.dhp.schema.oaf.Datasource;
-import eu.dnetlib.dhp.schema.oaf.ExternalReference;
-import eu.dnetlib.dhp.schema.oaf.Field;
-import eu.dnetlib.dhp.schema.oaf.Instance;
-import eu.dnetlib.dhp.schema.oaf.Journal;
-import eu.dnetlib.dhp.schema.oaf.KeyValue;
-import eu.dnetlib.dhp.schema.oaf.Project;
-import eu.dnetlib.dhp.schema.oaf.Publication;
-import eu.dnetlib.dhp.schema.oaf.Qualifier;
-import eu.dnetlib.dhp.schema.oaf.Result;
-import eu.dnetlib.dhp.schema.oaf.Software;
-import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
 
 public class ConversionUtils {
 
@@ -71,6 +58,10 @@ public class ConversionUtils {
 		return sp != null ? new OaBrokerTypedValue(classId(sp.getQualifier()), sp.getValue()) : null;
 	}
 
+	public static OaBrokerTypedValue oafSubjectToBrokerTypedValue(final Subject sp) {
+		return sp != null ? new OaBrokerTypedValue(classId(sp.getQualifier()), sp.getValue()) : null;
+	}
+
 	public static OaBrokerRelatedDataset oafDatasetToBrokerDataset(final Dataset d) {
 		if (d == null) {
 			return null;
@@ -115,7 +106,7 @@ public class ConversionUtils {
 		res.setTitles(structPropList(result.getTitle()));
 		res.setAbstracts(fieldList(result.getDescription()));
 		res.setLanguage(classId(result.getLanguage()));
-		res.setSubjects(structPropTypedList(result.getSubject()));
+		res.setSubjects(subjectList(result.getSubject()));
 		res.setCreators(mappedList(result.getAuthor(), ConversionUtils::oafAuthorToBrokerAuthor));
 		res.setPublicationdate(fieldValue(result.getDateofacceptance()));
 		res.setPublisher(fieldValue(result.getPublisher()));
@@ -304,6 +295,18 @@ public class ConversionUtils {
 			.collect(Collectors.toList());
 	}
 
+	private static List<OaBrokerTypedValue> subjectList(final List<Subject> list) {
+		if (list == null) {
+			return new ArrayList<>();
+		}
+
+		return list
+			.stream()
+			.map(ConversionUtils::oafSubjectToBrokerTypedValue)
+			.filter(Objects::nonNull)
+			.collect(Collectors.toList());
+	}
+
 	private static <F, T> List<T> mappedList(final List<F> list, final Function<F, T> func) {
 		if (list == null) {
 			return new ArrayList<>();

dhp-workflows/dhp-doiboost/src/main/scala/eu/dnetlib/doiboost/DoiBoostMappingUtil.scala

@@ -391,6 +391,28 @@ object DoiBoostMappingUtil {
     di
   }
 
+  def createSubject(value: String, classId: String, schemeId: String): Subject = {
+    val s = new Subject
+    s.setQualifier(OafMapperUtils.qualifier(classId, classId, schemeId, schemeId))
+    s.setValue(value)
+    s
+
+  }
+
+  def createSubject(
+    value: String,
+    classId: String,
+    className: String,
+    schemeId: String,
+    schemeName: String
+  ): Subject = {
+    val s = new Subject
+    s.setQualifier(OafMapperUtils.qualifier(classId, className, schemeId, schemeName))
+    s.setValue(value)
+    s
+
+  }
+
   def createSP(
     value: String,
     classId: String,

dhp-workflows/dhp-doiboost/src/main/scala/eu/dnetlib/doiboost/crossref/Crossref2Oaf.scala

@@ -201,7 +201,7 @@ case object Crossref2Oaf {
 
     if (subjectList.nonEmpty) {
       result.setSubject(
-        subjectList.map(s => createSP(s, "keywords", ModelConstants.DNET_SUBJECT_TYPOLOGIES)).asJava
+        subjectList.map(s => createSubject(s, "keyword", ModelConstants.DNET_SUBJECT_TYPOLOGIES)).asJava
       )
     }
 

dhp-workflows/dhp-doiboost/src/main/scala/eu/dnetlib/doiboost/mag/MagDataModel.scala

@@ -2,7 +2,7 @@ package eu.dnetlib.doiboost.mag
 
 import eu.dnetlib.dhp.schema.common.ModelConstants
 import eu.dnetlib.dhp.schema.oaf.utils.IdentifierFactory
-import eu.dnetlib.dhp.schema.oaf.{Instance, Journal, Publication, StructuredProperty}
+import eu.dnetlib.dhp.schema.oaf.{Instance, Journal, Publication, StructuredProperty, Subject}
 import eu.dnetlib.doiboost.DoiBoostMappingUtil
 import eu.dnetlib.doiboost.DoiBoostMappingUtil._
 import org.json4s
@@ -210,8 +210,8 @@ case object ConversionUtil {
       val className = "Microsoft Academic Graph classification"
       val classid = "MAG"
 
-      val p: List[StructuredProperty] = fieldOfStudy.subjects.flatMap(s => {
+      val p: List[Subject] = fieldOfStudy.subjects.flatMap(s => {
-        val s1 = createSP(
+        val s1 = createSubject(
           s.DisplayName,
           classid,
           className,
@@ -219,10 +219,10 @@ case object ConversionUtil {
           ModelConstants.DNET_SUBJECT_TYPOLOGIES
         )
         val di = DoiBoostMappingUtil.generateDataInfo(s.Score.toString)
-        var resList: List[StructuredProperty] = List(s1)
+        var resList: List[Subject] = List(s1)
         if (s.MainType.isDefined) {
           val maintp = s.MainType.get
-          val s2 = createSP(
+          val s2 = createSubject(
             s.MainType.get,
             classid,
             className,
@@ -232,7 +232,7 @@ case object ConversionUtil {
           s2.setDataInfo(di)
           resList = resList ::: List(s2)
           if (maintp.contains(".")) {
-            val s3 = createSP(
+            val s3 = createSubject(
               maintp.split("\\.").head,
               classid,
               className,

dhp-workflows/dhp-enrichment/src/test/java/eu/dnetlib/dhp/bulktag/EOSCTagJobTest.java

@@ -290,7 +290,7 @@ public class EOSCTagJobTest {
 					.stream()
 					.anyMatch(s -> s.getCode().equals("EOSC::Jupyter Notebook")));
 
-		List<StructuredProperty> subjects = tmp
+		List<Subject> subjects = tmp
 			.filter(sw -> sw.getId().equals("50|od______1582::6e7a9b21a2feef45673890432af34244"))
 			.collect()
 			.get(0)

dhp-workflows/dhp-graph-mapper/src/main/java/eu/dnetlib/dhp/oa/graph/clean/CleaningRuleMap.java

@@ -3,16 +3,15 @@ package eu.dnetlib.dhp.oa.graph.clean;
 
 import java.io.Serializable;
 import java.util.HashMap;
+import java.util.concurrent.atomic.AtomicReference;
 
+import org.apache.commons.lang3.SerializationUtils;
 import org.apache.commons.lang3.StringUtils;
 
 import eu.dnetlib.dhp.common.FunctionalInterfaceSupport.SerializableConsumer;
 import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
-import eu.dnetlib.dhp.schema.oaf.AccessRight;
+import eu.dnetlib.dhp.schema.oaf.*;
-import eu.dnetlib.dhp.schema.oaf.Country;
-import eu.dnetlib.dhp.schema.oaf.Qualifier;
-import eu.dnetlib.dhp.schema.oaf.Relation;
 
 public class CleaningRuleMap extends HashMap<Class<?>, SerializableConsumer<Object>> implements Serializable {
 
@@ -27,9 +26,36 @@ public class CleaningRuleMap extends HashMap<Class<?>, SerializableConsumer<Obje
 		mapping.put(AccessRight.class, o -> cleanQualifier(vocabularies, (AccessRight) o));
 		mapping.put(Country.class, o -> cleanCountry(vocabularies, (Country) o));
 		mapping.put(Relation.class, o -> cleanRelation(vocabularies, (Relation) o));
+		mapping.put(Subject.class, o -> cleanSubject(vocabularies, (Subject) o));
 		return mapping;
 	}
 
+	private static void cleanSubject(VocabularyGroup vocabularies, Subject subject) {
+		if (cleanSubjectForVocabulary(ModelConstants.DNET_SUBJECT_FOS_CLASSID, vocabularies, subject)) {
+			return;
+		} else {
+			// TODO cleaning based on different subject vocabs can be added here
+		}
+	}
+
+	private static boolean cleanSubjectForVocabulary(String vocabularyId, VocabularyGroup vocabularies,
+		Subject subject) {
+		AtomicReference<Boolean> modified = new AtomicReference<>(false);
+		vocabularies.find(vocabularyId).ifPresent(vocabulary -> {
+			if (!ModelConstants.DNET_SUBJECT_KEYWORD.equalsIgnoreCase(subject.getQualifier().getClassid())) {
+				return;
+			}
+			Qualifier newValue = vocabulary.lookup(subject.getValue());
+			if (!ModelConstants.UNKNOWN.equals(newValue.getClassid())) {
+				subject.setValue(newValue.getClassid());
+				subject.getQualifier().setClassid(vocabularyId);
+				subject.getQualifier().setClassname(vocabulary.getName());
+				modified.set(true);
+			}
+		});
+		return modified.get();
+	}
+
 	private static void cleanRelation(VocabularyGroup vocabularies, Relation r) {
 		if (vocabularies.vocabularyExists(ModelConstants.DNET_RELATION_SUBRELTYPE)) {
 			Qualifier newValue = vocabularies.lookup(ModelConstants.DNET_RELATION_SUBRELTYPE, r.getSubRelType());

dhp-workflows/dhp-graph-mapper/src/main/java/eu/dnetlib/dhp/oa/graph/raw/AbstractMdRecordToOafMapper.java

@@ -8,15 +8,7 @@ import static eu.dnetlib.dhp.schema.common.ModelConstants.PRODUCES;
 import static eu.dnetlib.dhp.schema.common.ModelConstants.REPOSITORY_PROVENANCE_ACTIONS;
 import static eu.dnetlib.dhp.schema.common.ModelConstants.RESULT_PROJECT;
 import static eu.dnetlib.dhp.schema.common.ModelConstants.UNKNOWN;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.createOpenaireId;
+import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.*;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.dataInfo;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.field;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.journal;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.keyValue;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.listFields;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.oaiIProvenance;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.qualifier;
-import static eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils.structuredProperty;
 
 import java.util.*;
 
@@ -29,26 +21,7 @@ import com.google.common.collect.Sets;
 import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
 import eu.dnetlib.dhp.schema.common.ModelConstants;
 import eu.dnetlib.dhp.schema.common.ModelSupport;
-import eu.dnetlib.dhp.schema.oaf.AccessRight;
+import eu.dnetlib.dhp.schema.oaf.*;
-import eu.dnetlib.dhp.schema.oaf.Author;
-import eu.dnetlib.dhp.schema.oaf.Context;
-import eu.dnetlib.dhp.schema.oaf.DataInfo;
-import eu.dnetlib.dhp.schema.oaf.Dataset;
-import eu.dnetlib.dhp.schema.oaf.Field;
-import eu.dnetlib.dhp.schema.oaf.GeoLocation;
-import eu.dnetlib.dhp.schema.oaf.Instance;
-import eu.dnetlib.dhp.schema.oaf.Journal;
-import eu.dnetlib.dhp.schema.oaf.KeyValue;
-import eu.dnetlib.dhp.schema.oaf.OAIProvenance;
-import eu.dnetlib.dhp.schema.oaf.Oaf;
-import eu.dnetlib.dhp.schema.oaf.OafEntity;
-import eu.dnetlib.dhp.schema.oaf.OtherResearchProduct;
-import eu.dnetlib.dhp.schema.oaf.Publication;
-import eu.dnetlib.dhp.schema.oaf.Qualifier;
-import eu.dnetlib.dhp.schema.oaf.Relation;
-import eu.dnetlib.dhp.schema.oaf.Result;
-import eu.dnetlib.dhp.schema.oaf.Software;
-import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
 import eu.dnetlib.dhp.schema.oaf.utils.IdentifierFactory;
 import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
 
@@ -411,7 +384,7 @@ public abstract class AbstractMdRecordToOafMapper {
 
 	protected abstract List<StructuredProperty> prepareTitles(Document doc, DataInfo info);
 
-	protected abstract List<StructuredProperty> prepareSubjects(Document doc, DataInfo info);
+	protected abstract List<Subject> prepareSubjects(Document doc, DataInfo info);
 
 	protected abstract Qualifier prepareLanguages(Document doc);
 
@@ -559,6 +532,22 @@ public abstract class AbstractMdRecordToOafMapper {
 		return res;
 	}
 
+	protected List<Subject> prepareSubjectList(
+		final Node node,
+		final String xpath,
+		final DataInfo info) {
+		final List<Subject> res = new ArrayList<>();
+		for (final Object o : node.selectNodes(xpath)) {
+			final Node n = (Node) o;
+			res
+				.add(
+					subject(
+						n.getText(), n.valueOf("@classid"), n.valueOf("@classname"), n.valueOf("@schemeid"),
+						n.valueOf("@schemename"), info));
+		}
+		return res;
+	}
+
 	protected OAIProvenance prepareOAIprovenance(final Document doc) {
 		final Node n = doc.selectSingleNode("//*[local-name()='provenance']/*[local-name()='originDescription']");
 

dhp-workflows/dhp-graph-mapper/src/main/java/eu/dnetlib/dhp/oa/graph/raw/OafToOafMapper.java

@@ -84,8 +84,8 @@ public class OafToOafMapper extends AbstractMdRecordToOafMapper {
 	}
 
 	@Override
-	protected List<StructuredProperty> prepareSubjects(final Document doc, final DataInfo info) {
+	protected List<Subject> prepareSubjects(final Document doc, final DataInfo info) {
-		return prepareListStructProps(doc, "//dc:subject", info);
+		return prepareSubjectList(doc, "//dc:subject", info);
 	}
 
 	@Override

dhp-workflows/dhp-graph-mapper/src/main/java/eu/dnetlib/dhp/oa/graph/raw/OdfToOafMapper.java

@@ -257,8 +257,8 @@ public class OdfToOafMapper extends AbstractMdRecordToOafMapper {
 	}
 
 	@Override
-	protected List<StructuredProperty> prepareSubjects(final Document doc, final DataInfo info) {
+	protected List<Subject> prepareSubjects(final Document doc, final DataInfo info) {
-		return prepareListStructProps(doc, "//*[local-name()='subject']", info);
+		return prepareSubjectList(doc, "//*[local-name()='subject']", info);
 	}
 
 	@Override

dhp-workflows/dhp-graph-mapper/src/test/java/eu/dnetlib/dhp/oa/graph/clean/GraphCleaningFunctionsTest.java

@@ -251,6 +251,33 @@ public class GraphCleaningFunctionsTest {
 							pid.getQualifier().getClassname()));
 		});
 
+		assertNotNull(p_cleaned.getSubject());
+
+		List<Subject> fos_subjects = p_cleaned
+			.getSubject()
+			.stream()
+			.filter(s -> ModelConstants.DNET_SUBJECT_FOS_CLASSID.equals(s.getQualifier().getClassid()))
+			.collect(Collectors.toList());
+
+		assertNotNull(fos_subjects);
+		assertEquals(2, fos_subjects.size());
+
+		assertTrue(
+			fos_subjects
+				.stream()
+				.anyMatch(
+					s -> "0101 mathematics".equals(s.getValue()) &
+						ModelConstants.DNET_SUBJECT_FOS_CLASSID.equals(s.getQualifier().getClassid()) &
+						"sysimport:crosswalk:datasetarchive"
+							.equals(s.getDataInfo().getProvenanceaction().getClassid())));
+
+		assertTrue(
+			fos_subjects
+				.stream()
+				.anyMatch(
+					s -> "0102 computer and information sciences".equals(s.getValue()) &
+						ModelConstants.DNET_SUBJECT_FOS_CLASSID.equals(s.getQualifier().getClassid())));
+
 		// TODO add more assertions to verity the cleaned values
 		System.out.println(MAPPER.writeValueAsString(p_cleaned));
 	}

dhp-workflows/dhp-graph-mapper/src/test/java/eu/dnetlib/dhp/oa/graph/raw/MappersTest.java

@@ -928,7 +928,8 @@ class MappersTest {
 
 	@Test
 	void testROHub2() throws IOException, DocumentException {
-		final String xml = IOUtils.toString(Objects.requireNonNull(getClass().getResourceAsStream("rohub-modified.xml")));
+		final String xml = IOUtils
+			.toString(Objects.requireNonNull(getClass().getResourceAsStream("rohub-modified.xml")));
 		final List<Oaf> list = new OdfToOafMapper(vocs, false, true).processMdRecord(xml);
 		System.out.println("***************");
 		System.out.println(new ObjectMapper().writeValueAsString(list));

dhp-workflows/dhp-graph-mapper/src/test/resources/eu/dnetlib/dhp/oa/graph/clean/result.json

@@ -743,12 +743,12 @@
         "trust": "0.9"
       },
       "qualifier": {
-        "classid": "",
+        "classid": "keyword",
-        "classname": "",
+        "classname": "keyword",
-        "schemeid": "",
+        "schemeid": "dnet:subject_classification_typologies",
-        "schemename": ""
+        "schemename": "dnet:subject_classification_typologies"
       },
-      "value": "infrared detectors"
+      "value": "FOS: Mathematics"
     },
     {
       "dataInfo": {
@@ -765,12 +765,12 @@
         "trust": "0.9"
       },
       "qualifier": {
-        "classid": "",
+        "classid": "keyword",
-        "classname": "",
+        "classname": "keyword",
-        "schemeid": "",
+        "schemeid": "dnet:subject_classification_typologies",
-        "schemename": ""
+        "schemename": "dnet:subject_classification_typologies"
       },
-      "value": "lens antennas"
+      "value": "FOS: Computer and information sciences"
     },
     {
       "dataInfo": {
@@ -787,12 +787,34 @@
         "trust": "0.9"
       },
       "qualifier": {
-        "classid": "",
+        "classid": "keyword",
-        "classname": "",
+        "classname": "keyword",
-        "schemeid": "",
+        "schemeid": "dnet:subject_classification_typologies",
-        "schemename": ""
+        "schemename": "dnet:subject_classification_typologies"
       },
-      "value": "silicon"
+      "value": "0101 mathematics"
+    },
+    {
+      "dataInfo": {
+        "deletedbyinference": false,
+        "inferenceprovenance": "",
+        "inferred": false,
+        "invisible": false,
+        "provenanceaction": {
+          "classid": "subject:fos",
+          "classname": "subject:fos",
+          "schemeid": "dnet:provenanceActions",
+          "schemename": "dnet:provenanceActions"
+        },
+        "trust": "0.9"
+      },
+      "qualifier": {
+        "classid": "keyword",
+        "classname": "keyword",
+        "schemeid": "dnet:subject_classification_typologies",
+        "schemename": "dnet:subject_classification_typologies"
+      },
+      "value": "0101 mathematics"
     },
     {
       "dataInfo": {

dhp-workflows/dhp-graph-mapper/src/test/resources/eu/dnetlib/dhp/oa/graph/clean/synonyms.txt

@@ -1244,3 +1244,5 @@ dnet:relation_relClass @=@ IsSupplementedBy @=@ isSupplementedBy
 dnet:relation_relClass @=@ IsRelatedTo @=@ isRelatedTo
 dnet:relation_subRelType @=@ relationship @=@ publicationDataset
 dnet:provenanceActions @=@ iis @=@ erroneous label to be cleaned
+FOS @=@ 0101 mathematics @=@ FOS: Mathematics
+FOS @=@ 0102 computer and information sciences @=@ FOS: Computer and information sciences

dhp-workflows/dhp-graph-mapper/src/test/resources/eu/dnetlib/dhp/oa/graph/clean/terms.txt

@@ -1119,3 +1119,6 @@ dnet:relation_subRelType @=@ dnet:relation_subRelType @=@ review @=@ review
 dnet:relation_subRelType @=@ dnet:relation_subRelType @=@ similarity @=@ similarity
 dnet:relation_subRelType @=@ dnet:relation_subRelType @=@ supplement @=@ supplement
 dnet:relation_subRelType @=@ dnet:relation_subRelType @=@ version @=@ version
+FOS @=@ Fields of Science and Technology classification @=@ 0101 mathematics @=@ 0101 mathematics
+FOS @=@ Fields of Science and Technology classification @=@ 0102 computer and information sciences @=@ 0102 computer and information sciences
+FOS @=@ Fields of Science and Technology classification @=@ 0103 physical sciences @=@ 0103 physical sciences

dhp-workflows/dhp-graph-mapper/src/test/scala/eu/dnetlib/dhp/oa/graph/resolution/ResolveEntitiesTest.scala

@@ -59,7 +59,7 @@ class ResolveEntitiesTest extends Serializable {
         r.setId(id.toLowerCase.trim)
         r.setSubject(
           List(
-            OafMapperUtils.structuredProperty(
+            OafMapperUtils.subject(
               FAKE_SUBJECT,
               OafMapperUtils.qualifier("fos", "fosCS", "fossSchema", "fossiFIgo"),
               null
@@ -250,7 +250,7 @@ class ResolveEntitiesTest extends Serializable {
     val r = new Result
     r.setSubject(
       List(
-        OafMapperUtils.structuredProperty(
+        OafMapperUtils.subject(
           FAKE_SUBJECT,
           OafMapperUtils.qualifier("fos", "fosCS", "fossSchema", "fossiFIgo"),
           null

dhp-workflows/dhp-graph-provision/src/test/java/eu/dnetlib/dhp/oa/provision/IndexRecordTransformerTest.java

@@ -95,7 +95,6 @@ public class IndexRecordTransformerTest {
 		testRecordTransformation(record);
 	}
 
-
 	@Test
 	public void testForEOSCFutureDataTransferPilot() throws IOException, TransformerException {
 		final String record = IOUtils.toString(getClass().getResourceAsStream("eosc-future/data-transfer-pilot.xml"));
@@ -129,8 +128,6 @@ public class IndexRecordTransformerTest {
 		testRecordTransformation(record);
 	}
 
-
-
 	@Test
 	void testDoiUrlNormalization() throws MalformedURLException {
 

pom.xml

@@ -801,7 +801,7 @@
 		<mockito-core.version>3.3.3</mockito-core.version>
 		<mongodb.driver.version>3.4.2</mongodb.driver.version>
 		<vtd.version>[2.12,3.0)</vtd.version>
-		<dhp-schemas.version>[2.12.2-SNAPSHOT]</dhp-schemas.version>
+		<dhp-schemas.version>[3.14.0]</dhp-schemas.version>
 		<dnet-actionmanager-api.version>[4.0.3]</dnet-actionmanager-api.version>
 		<dnet-actionmanager-common.version>[6.0.5]</dnet-actionmanager-common.version>
 		<dnet-openaire-broker-common.version>[3.1.6]</dnet-openaire-broker-common.version>