sabeel
/
dnet-hadoop
forked from D-Net/dnet-hadoop
1
0
Fork 0
Code Pull Requests Activity

adjusted dedup configuration for result entities: using new wordssuffixprefix clustering function, removed ngrampairs, adjusted queueMaxSize (800) and slidingWindowSize (80)

This commit is contained in:
Claudio Atzori 2020-07-02 17:35:22 +02:00
parent 1d39f7901c
commit c3d67f709a
6 changed files with 16 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
dhp-workflows/dhp-dedup-openaire/src/test/java/eu/dnetlib/dhp/oa/dedup/SparkDedupTest.java
View File

@ -182,7 +182,7 @@ public class SparkDedupTest implements Serializable {
.count(); .count();
assertEquals(3432, orgs_simrel); assertEquals(3432, orgs_simrel);
assertEquals(7152, pubs_simrel); assertEquals(7054, pubs_simrel);
assertEquals(344, sw_simrel); assertEquals(344, sw_simrel);
assertEquals(458, ds_simrel); assertEquals(458, ds_simrel);
assertEquals(6750, orp_simrel); assertEquals(6750, orp_simrel);
@ -234,7 +234,7 @@ public class SparkDedupTest implements Serializable {
.count(); .count();
assertEquals(1276, orgs_mergerel); assertEquals(1276, orgs_mergerel);
assertEquals(1442, pubs_mergerel); assertEquals(1440, pubs_mergerel);
assertEquals(288, sw_mergerel); assertEquals(288, sw_mergerel);
assertEquals(472, ds_mergerel); assertEquals(472, ds_mergerel);
assertEquals(718, orp_mergerel); assertEquals(718, orp_mergerel);
@ -423,7 +423,7 @@ public class SparkDedupTest implements Serializable {
long relations = jsc.textFile(testDedupGraphBasePath + "/relation").count(); long relations = jsc.textFile(testDedupGraphBasePath + "/relation").count();
assertEquals(4975, relations); assertEquals(4971, relations);
// check deletedbyinference // check deletedbyinference
final Dataset<Relation> mergeRels = spark final Dataset<Relation> mergeRels = spark

7
dhp-workflows/dhp-dedup-openaire/src/test/resources/eu/dnetlib/dhp/dedup/conf/ds.curr.conf.json
View File

@ -6,10 +6,10 @@
"subEntityType" : "resulttype", "subEntityType" : "resulttype",
"subEntityValue" : "dataset", "subEntityValue" : "dataset",
"orderField" : "title", "orderField" : "title",
"queueMaxSize" : "2000", "queueMaxSize" : "800",
"groupMaxSize" : "100", "groupMaxSize" : "100",
"maxChildren" : "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" ], "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", "includeChildren" : "true",
"idPath" : "$.id", "idPath" : "$.id",
@ -17,8 +17,7 @@
}, },
"pace" : { "pace" : {
"clustering" : [ "clustering" : [
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} }, { "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } } { "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
], ],
"decisionTree" : { "decisionTree" : {

7
dhp-workflows/dhp-dedup-openaire/src/test/resources/eu/dnetlib/dhp/dedup/conf/orp.curr.conf.json
View File

@ -6,10 +6,10 @@
"subEntityType" : "resulttype", "subEntityType" : "resulttype",
"subEntityValue" : "otherresearchproduct", "subEntityValue" : "otherresearchproduct",
"orderField" : "title", "orderField" : "title",
"queueMaxSize" : "2000", "queueMaxSize" : "800",
"groupMaxSize" : "100", "groupMaxSize" : "100",
"maxChildren" : "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" ], "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", "includeChildren" : "true",
"idPath" : "$.id", "idPath" : "$.id",
@ -17,8 +17,7 @@
}, },
"pace" : { "pace" : {
"clustering" : [ "clustering" : [
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} }, { "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } } { "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
], ],
"decisionTree" : { "decisionTree" : {

7
dhp-workflows/dhp-dedup-openaire/src/test/resources/eu/dnetlib/dhp/dedup/conf/pub.curr.conf.json
View File

@ -6,10 +6,10 @@
"subEntityType": "resulttype", "subEntityType": "resulttype",
"subEntityValue": "publication", "subEntityValue": "publication",
"orderField": "title", "orderField": "title",
"queueMaxSize": "2000", "queueMaxSize": "800",
"groupMaxSize": "100", "groupMaxSize": "100",
"maxChildren": "100", "maxChildren": "100",
"slidingWindowSize": "200", "slidingWindowSize": "80",
"rootBuilder": [ "rootBuilder": [
"result", "result",
"resultProject_outcome_isProducedBy", "resultProject_outcome_isProducedBy",
@ -29,8 +29,7 @@
}, },
"pace": { "pace": {
"clustering" : [ "clustering" : [
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} }, { "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
{ "name" : "lowercase", "fields" : [ "doi" ], "params" : { } } { "name" : "lowercase", "fields" : [ "doi" ], "params" : { } }
], ],
"decisionTree": { "decisionTree": {

7
dhp-workflows/dhp-dedup-openaire/src/test/resources/eu/dnetlib/dhp/dedup/conf/sw.curr.conf.json
View File

@ -6,10 +6,10 @@
"subEntityType" : "resulttype", "subEntityType" : "resulttype",
"subEntityValue" : "software", "subEntityValue" : "software",
"orderField" : "title", "orderField" : "title",
"queueMaxSize" : "2000", "queueMaxSize" : "800",
"groupMaxSize" : "100", "groupMaxSize" : "100",
"maxChildren" : "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" ], "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", "includeChildren" : "true",
"idPath" : "$.id", "idPath" : "$.id",
@ -17,8 +17,7 @@
}, },
"pace" : { "pace" : {
"clustering" : [ "clustering" : [
{ "name" : "ngrampairs", "fields" : [ "title" ], "params" : { "max" : "1", "ngramLen" : "3"} }, { "name" : "wordssuffixprefix", "fields" : [ "title" ], "params" : { "max" : "2", "len" : "3" } },
{ "name" : "suffixprefix", "fields" : [ "title" ], "params" : { "max" : "1", "len" : "3" } },
{ "name" : "lowercase", "fields" : [ "doi", "url" ], "params" : { } } { "name" : "lowercase", "fields" : [ "doi", "url" ], "params" : { } }
], ],
"decisionTree": { "decisionTree": {

2
pom.xml
View File

@ -315,7 +315,7 @@
<dependency> <dependency>
<groupId>eu.dnetlib</groupId> <groupId>eu.dnetlib</groupId>
<artifactId>dnet-pace-core</artifactId> <artifactId>dnet-pace-core</artifactId>
<version>4.0.1</version> <version>4.0.2</version>
</dependency> </dependency>
<dependency> <dependency>
<groupId>eu.dnetlib</groupId> <groupId>eu.dnetlib</groupId>