forked from D-Net/dnet-hadoop
Merge branch 'master' of code-repo.d4science.org:D-Net/dnet-hadoop
This commit is contained in:
commit
dffa0b01a2
|
@ -182,7 +182,7 @@ public class SparkDedupTest implements Serializable {
|
|||
.count();
|
||||
|
||||
assertEquals(3432, orgs_simrel);
|
||||
assertEquals(7152, pubs_simrel);
|
||||
assertEquals(7054, pubs_simrel);
|
||||
assertEquals(344, sw_simrel);
|
||||
assertEquals(458, ds_simrel);
|
||||
assertEquals(6750, orp_simrel);
|
||||
|
@ -234,7 +234,7 @@ public class SparkDedupTest implements Serializable {
|
|||
.count();
|
||||
|
||||
assertEquals(1276, orgs_mergerel);
|
||||
assertEquals(1442, pubs_mergerel);
|
||||
assertEquals(1440, pubs_mergerel);
|
||||
assertEquals(288, sw_mergerel);
|
||||
assertEquals(472, ds_mergerel);
|
||||
assertEquals(718, orp_mergerel);
|
||||
|
@ -423,7 +423,7 @@ public class SparkDedupTest implements Serializable {
|
|||
|
||||
long relations = jsc.textFile(testDedupGraphBasePath + "/relation").count();
|
||||
|
||||
assertEquals(4975, relations);
|
||||
assertEquals(4971, relations);
|
||||
|
||||
// check deletedbyinference
|
||||
final Dataset<Relation> mergeRels = spark
|
||||
|
|
|
@ -6,10 +6,10 @@
|
|||
"subEntityType" : "resulttype",
|
||||
"subEntityValue" : "dataset",
|
||||
"orderField" : "title",
|
||||
"queueMaxSize" : "2000",
|
||||
"queueMaxSize" : "800",
|
||||
"groupMaxSize" : "100",
|
||||
"maxChildren" : "100",
|
||||
"slidingWindowSize" : "200",
|
||||
"slidingWindowSize" : "80",
|
||||
"rootBuilder" : ["result", "resultProject_outcome_isProducedBy", "resultResult_publicationDataset_isRelatedTo", "resultResult_similarity_isAmongTopNSimilarDocuments", "resultResult_similarity_hasAmongTopNSimilarDocuments", "resultOrganization_affiliation_hasAuthorInstitution", "resultResult_part_hasPart", "resultResult_part_isPartOf", "resultResult_supplement_isSupplementTo", "resultResult_supplement_isSupplementedBy", "resultResult_version_isVersionOf" ],
|
||||
"includeChildren" : "true",
|
||||
"idPath" : "$.id",
|
||||
|
@ -17,8 +17,7 @@
|
|||
},
|
||||
"pace" : {
|
||||
"clustering" : [
|
||||
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} },
|
||||
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
|
||||
{ "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
|
||||
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
|
||||
],
|
||||
"decisionTree" : {
|
||||
|
|
|
@ -6,10 +6,10 @@
|
|||
"subEntityType" : "resulttype",
|
||||
"subEntityValue" : "otherresearchproduct",
|
||||
"orderField" : "title",
|
||||
"queueMaxSize" : "2000",
|
||||
"queueMaxSize" : "800",
|
||||
"groupMaxSize" : "100",
|
||||
"maxChildren" : "100",
|
||||
"slidingWindowSize" : "200",
|
||||
"slidingWindowSize" : "80",
|
||||
"rootBuilder" : [ "result", "resultProject_outcome_isProducedBy", "resultResult_publicationDataset_isRelatedTo", "resultResult_similarity_isAmongTopNSimilarDocuments", "resultResult_similarity_hasAmongTopNSimilarDocuments", "resultOrganization_affiliation_hasAuthorInstitution", "resultResult_part_hasPart", "resultResult_part_isPartOf", "resultResult_supplement_isSupplementTo", "resultResult_supplement_isSupplementedBy", "resultResult_version_isVersionOf" ],
|
||||
"includeChildren" : "true",
|
||||
"idPath" : "$.id",
|
||||
|
@ -17,8 +17,7 @@
|
|||
},
|
||||
"pace" : {
|
||||
"clustering" : [
|
||||
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} },
|
||||
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
|
||||
{ "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
|
||||
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
|
||||
],
|
||||
"decisionTree" : {
|
||||
|
|
|
@ -6,10 +6,10 @@
|
|||
"subEntityType": "resulttype",
|
||||
"subEntityValue": "publication",
|
||||
"orderField": "title",
|
||||
"queueMaxSize": "2000",
|
||||
"queueMaxSize": "800",
|
||||
"groupMaxSize": "100",
|
||||
"maxChildren": "100",
|
||||
"slidingWindowSize": "200",
|
||||
"slidingWindowSize": "80",
|
||||
"rootBuilder": [
|
||||
"result",
|
||||
"resultProject_outcome_isProducedBy",
|
||||
|
@ -29,8 +29,7 @@
|
|||
},
|
||||
"pace": {
|
||||
"clustering" : [
|
||||
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} },
|
||||
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
|
||||
{ "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
|
||||
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
|
||||
],
|
||||
"decisionTree": {
|
||||
|
|
|
@ -6,10 +6,10 @@
|
|||
"subEntityType" : "resulttype",
|
||||
"subEntityValue" : "software",
|
||||
"orderField" : "title",
|
||||
"queueMaxSize" : "2000",
|
||||
"queueMaxSize" : "800",
|
||||
"groupMaxSize" : "100",
|
||||
"maxChildren" : "100",
|
||||
"slidingWindowSize" : "200",
|
||||
"slidingWindowSize" : "80",
|
||||
"rootBuilder" : [ "result", "resultProject_outcome_isProducedBy", "resultResult_publicationDataset_isRelatedTo", "resultResult_similarity_isAmongTopNSimilarDocuments", "resultResult_similarity_hasAmongTopNSimilarDocuments", "resultOrganization_affiliation_hasAuthorInstitution", "resultResult_part_hasPart", "resultResult_part_isPartOf", "resultResult_supplement_isSupplementTo", "resultResult_supplement_isSupplementedBy", "resultResult_version_isVersionOf" ],
|
||||
"includeChildren" : "true",
|
||||
"idPath" : "$.id",
|
||||
|
@ -17,8 +17,7 @@
|
|||
},
|
||||
"pace" : {
|
||||
"clustering" : [
|
||||
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} },
|
||||
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
|
||||
{ "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
|
||||
{ "name" : "lowercase", "fields" : [ "doi", "url" ], "params" : { } }
|
||||
],
|
||||
"decisionTree": {
|
||||
|
|
|
@ -156,6 +156,7 @@ public class PrepareRelationsJob {
|
|||
.parquet(outputPath);
|
||||
}
|
||||
|
||||
// experimental
|
||||
private static void prepareRelationsDataset(
|
||||
SparkSession spark, String inputRelationsPath, String outputPath, Set<String> relationFilter, int maxRelations,
|
||||
int relPartitions) {
|
||||
|
|
|
@ -134,6 +134,7 @@
|
|||
<arg>--inputRelationsPath</arg><arg>${inputGraphRootPath}/relation</arg>
|
||||
<arg>--outputPath</arg><arg>${workingDir}/relation</arg>
|
||||
<arg>--maxRelations</arg><arg>${maxRelations}</arg>
|
||||
<arg>--relationFilter</arg><arg>${relationFilter}</arg>
|
||||
<arg>--relPartitions</arg><arg>5000</arg>
|
||||
</spark>
|
||||
<ok to="fork_join_related_entities"/>
|
||||
|
|
Loading…
Reference in New Issue