DedupConfig as json file

This commit is contained in:
Michele Artini 2020-07-09 12:53:46 +02:00
parent a44b9b36b9
commit 2d742a84ae
11 changed files with 208 additions and 117 deletions

View File

@ -18,8 +18,6 @@ import org.apache.spark.util.LongAccumulator;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.broker.model.Event;
import eu.dnetlib.dhp.broker.oa.matchers.UpdateMatcher;
@ -27,9 +25,6 @@ import eu.dnetlib.dhp.broker.oa.util.ClusterUtils;
import eu.dnetlib.dhp.broker.oa.util.EventFinder;
import eu.dnetlib.dhp.broker.oa.util.EventGroup;
import eu.dnetlib.dhp.broker.oa.util.aggregators.simple.ResultGroup;
import eu.dnetlib.dhp.utils.ISLookupClientFactory;
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
import eu.dnetlib.pace.config.DedupConfig;
public class GenerateEventsJob {
@ -52,12 +47,6 @@ public class GenerateEventsJob {
final String workingPath = parser.get("workingPath");
log.info("workingPath: {}", workingPath);
final String isLookupUrl = parser.get("isLookupUrl");
log.info("isLookupUrl: {}", isLookupUrl);
final String dedupConfigProfileId = parser.get("dedupConfProfile");
log.info("dedupConfigProfileId: {}", dedupConfigProfileId);
final String eventsPath = workingPath + "/events";
log.info("eventsPath: {}", eventsPath);
@ -72,10 +61,6 @@ public class GenerateEventsJob {
final SparkConf conf = new SparkConf();
// TODO UNCOMMENT
// final DedupConfig dedupConfig = loadDedupConfig(isLookupUrl, dedupConfigProfileId);
final DedupConfig dedupConfig = null;
runWithSparkSession(conf, isSparkSessionManaged, spark -> {
ClusterUtils.removeDir(spark, eventsPath);
@ -90,7 +75,7 @@ public class GenerateEventsJob {
final Dataset<Event> dataset = groups
.map(
g -> EventFinder
.generateEvents(g, dsIdWhitelist, dsIdBlacklist, dsTypeWhitelist, dedupConfig, accumulators),
.generateEvents(g, dsIdWhitelist, dsIdBlacklist, dsTypeWhitelist, accumulators),
Encoders
.bean(EventGroup.class))
.flatMap(g -> g.getData().iterator(), Encoders.bean(Event.class));
@ -112,23 +97,4 @@ public class GenerateEventsJob {
}
private static DedupConfig loadDedupConfig(final String isLookupUrl, final String profId) throws Exception {
final ISLookUpService isLookUpService = ISLookupClientFactory.getLookUpService(isLookupUrl);
final String conf = isLookUpService
.getResourceProfileByQuery(
String
.format(
"for $x in /RESOURCE_PROFILE[.//RESOURCE_IDENTIFIER/@value = '%s'] return $x//DEDUPLICATION/text()",
profId));
final DedupConfig dedupConfig = new ObjectMapper().readValue(conf, DedupConfig.class);
dedupConfig.getPace().initModel();
dedupConfig.getPace().initTranslationMap();
// dedupConfig.getWf().setConfigurationId("???");
return dedupConfig;
}
}

View File

@ -17,7 +17,6 @@ import org.apache.spark.util.LongAccumulator;
import eu.dnetlib.broker.objects.OaBrokerMainEntity;
import eu.dnetlib.dhp.broker.model.Topic;
import eu.dnetlib.dhp.broker.oa.util.UpdateInfo;
import eu.dnetlib.pace.config.DedupConfig;
public abstract class UpdateMatcher<T> {
@ -37,7 +36,6 @@ public abstract class UpdateMatcher<T> {
public Collection<UpdateInfo<T>> searchUpdatesForRecord(final OaBrokerMainEntity res,
final Collection<OaBrokerMainEntity> others,
final DedupConfig dedupConfig,
final Map<String, LongAccumulator> accumulators) {
final Map<String, UpdateInfo<T>> infoMap = new HashMap<>();
@ -49,7 +47,7 @@ public abstract class UpdateMatcher<T> {
if (topic != null) {
final UpdateInfo<T> info = new UpdateInfo<>(topic, hl, source, res,
getCompileHighlightFunction(),
getHighlightToStringFunction(), dedupConfig);
getHighlightToStringFunction());
final String s = DigestUtils.md5Hex(info.getHighlightValueAsString());
if (!infoMap.containsKey(s) || infoMap.get(s).getTrust() < info.getTrust()) {

View File

@ -37,7 +37,6 @@ import eu.dnetlib.dhp.broker.oa.matchers.simple.EnrichMoreOpenAccess;
import eu.dnetlib.dhp.broker.oa.matchers.simple.EnrichMorePid;
import eu.dnetlib.dhp.broker.oa.matchers.simple.EnrichMoreSubject;
import eu.dnetlib.dhp.broker.oa.util.aggregators.simple.ResultGroup;
import eu.dnetlib.pace.config.DedupConfig;
public class EventFinder {
@ -76,7 +75,6 @@ public class EventFinder {
final Set<String> dsIdWhitelist,
final Set<String> dsIdBlacklist,
final Set<String> dsTypeWhitelist,
final DedupConfig dedupConfig,
final Map<String, LongAccumulator> accumulators) {
final List<UpdateInfo<?>> list = new ArrayList<>();
@ -84,7 +82,7 @@ public class EventFinder {
for (final OaBrokerMainEntity target : results.getData()) {
if (verifyTarget(target, dsIdWhitelist, dsIdBlacklist, dsTypeWhitelist)) {
for (final UpdateMatcher<?> matcher : matchers) {
list.addAll(matcher.searchUpdatesForRecord(target, results.getData(), dedupConfig, accumulators));
list.addAll(matcher.searchUpdatesForRecord(target, results.getData(), accumulators));
}
}
}

View File

@ -1,8 +1,62 @@
package eu.dnetlib.dhp.broker.oa.util;
import java.io.IOException;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.broker.objects.OaBrokerMainEntity;
import eu.dnetlib.pace.config.DedupConfig;
import eu.dnetlib.pace.model.MapDocument;
import eu.dnetlib.pace.tree.support.TreeProcessor;
import eu.dnetlib.pace.util.MapDocumentUtil;
public class TrustUtils {
private static final Logger log = LoggerFactory.getLogger(TrustUtils.class);
private static DedupConfig dedupConfig;
static {
final ObjectMapper mapper = new ObjectMapper();
try {
dedupConfig = mapper
.readValue(
DedupConfig.class.getResourceAsStream("/eu/dnetlib/dhp/broker/oa/dedupConfig/dedupConfig.json"),
DedupConfig.class);
} catch (final IOException e) {
log.error("Error loading dedupConfig, e");
}
}
protected static float calculateTrust(final OaBrokerMainEntity r1, final OaBrokerMainEntity r2) {
if (dedupConfig == null) {
return BrokerConstants.MIN_TRUST;
}
try {
final ObjectMapper objectMapper = new ObjectMapper();
final MapDocument doc1 = MapDocumentUtil
.asMapDocumentWithJPath(dedupConfig, objectMapper.writeValueAsString(r1));
final MapDocument doc2 = MapDocumentUtil
.asMapDocumentWithJPath(dedupConfig, objectMapper.writeValueAsString(r2));
final double score = new TreeProcessor(dedupConfig).computeScore(doc1, doc2);
final double threshold = dedupConfig.getWf().getThreshold();
return TrustUtils.rescale(score, threshold);
} catch (final Exception e) {
log.error("Error computing score between results", e);
return BrokerConstants.MIN_TRUST;
}
}
public static float rescale(final double score, final double threshold) {
if (score >= BrokerConstants.MAX_TRUST) {
return BrokerConstants.MAX_TRUST;

View File

@ -4,20 +4,11 @@ package eu.dnetlib.dhp.broker.oa.util;
import java.util.function.BiConsumer;
import java.util.function.Function;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.broker.objects.OaBrokerEventPayload;
import eu.dnetlib.broker.objects.OaBrokerInstance;
import eu.dnetlib.broker.objects.OaBrokerMainEntity;
import eu.dnetlib.broker.objects.OaBrokerProvenance;
import eu.dnetlib.dhp.broker.model.Topic;
import eu.dnetlib.pace.config.DedupConfig;
import eu.dnetlib.pace.model.MapDocument;
import eu.dnetlib.pace.tree.support.TreeProcessor;
import eu.dnetlib.pace.util.MapDocumentUtil;
public final class UpdateInfo<T> {
@ -35,20 +26,17 @@ public final class UpdateInfo<T> {
private final float trust;
private static final Logger log = LoggerFactory.getLogger(UpdateInfo.class);
public UpdateInfo(final Topic topic, final T highlightValue, final OaBrokerMainEntity source,
final OaBrokerMainEntity target,
final BiConsumer<OaBrokerMainEntity, T> compileHighlight,
final Function<T, String> highlightToString,
final DedupConfig dedupConfig) {
final Function<T, String> highlightToString) {
this.topic = topic;
this.highlightValue = highlightValue;
this.source = source;
this.target = target;
this.compileHighlight = compileHighlight;
this.highlightToString = highlightToString;
this.trust = calculateTrust(dedupConfig, source, target);
this.trust = TrustUtils.calculateTrust(source, target);
}
public T getHighlightValue() {
@ -63,31 +51,6 @@ public final class UpdateInfo<T> {
return target;
}
private float calculateTrust(final DedupConfig dedupConfig,
final OaBrokerMainEntity r1,
final OaBrokerMainEntity r2) {
if (dedupConfig == null) {
return BrokerConstants.MIN_TRUST;
}
try {
final ObjectMapper objectMapper = new ObjectMapper();
final MapDocument doc1 = MapDocumentUtil
.asMapDocumentWithJPath(dedupConfig, objectMapper.writeValueAsString(r1));
final MapDocument doc2 = MapDocumentUtil
.asMapDocumentWithJPath(dedupConfig, objectMapper.writeValueAsString(r2));
final double score = new TreeProcessor(dedupConfig).computeScore(doc1, doc2);
final double threshold = dedupConfig.getWf().getThreshold();
return TrustUtils.rescale(score, threshold);
} catch (final Exception e) {
log.error("Error computing score between results", e);
return BrokerConstants.MIN_TRUST;
}
}
protected Topic getTopic() {
return topic;
}

View File

@ -0,0 +1,122 @@
{
"wf": {
},
"pace": {
"clustering": [
{
"name": "wordssuffixprefix",
"fields": [
"title"
],
"params": {
"max": "2",
"len": "3"
}
},
{
"name": "lowercase",
"fields": [
"doi"
],
"params": {
}
}
],
"decisionTree": {
"start": {
"fields": [
{
"field": "doi",
"comparator": "exactMatch",
"weight": 1.0,
"countIfUndefined": "false",
"params": {
}
}
],
"threshold": 0.5,
"aggregation": "AVG",
"positive": "MATCH",
"negative": "layer1",
"undefined": "layer1",
"ignoreUndefined": "true"
},
"layer1": {
"fields": [
{
"field": "title",
"comparator": "titleVersionMatch",
"weight": 0.9,
"countIfUndefined": "false",
"params": {
}
},
{
"field": "authors",
"comparator": "sizeMatch",
"weight": 0.9,
"countIfUndefined": "false",
"params": {
}
}
],
"threshold": 0.5,
"aggregation": "AVG",
"positive": "MATCH",
"negative": "layer2",
"undefined": "layer2",
"ignoreUndefined": "true"
},
"layer2": {
"fields": [
{
"field": "title",
"comparator": "levensteinTitle",
"weight": 1.0,
"countIfUndefined": "true",
"params": {
}
}
],
"threshold": 0.99,
"aggregation": "AVG",
"positive": "MATCH",
"negative": "NO_MATCH",
"undefined": "NO_MATCH",
"ignoreUndefined": "true"
}
},
"model": [
{
"name": "doi",
"type": "String",
"path": "$.pids[?(@.type == 'doi')].value"
},
{
"name": "title",
"type": "String",
"path": "$.titles",
"length": 250,
"size": 5
},
{
"name": "authors",
"type": "List",
"path": "$.creators[*].fullname",
"size": 200
}
],
"blacklists": {
},
"synonyms": {
}
}
}

View File

@ -8,14 +8,6 @@
<property>
<name>workingPath</name>
<description>the path where the the generated data will be stored</description>
</property>
<property>
<name>isLookupUrl</name>
<description>the address of the lookUp service</description>
</property>
<property>
<name>dedupConfProfId</name>
<description>the id of a valid Dedup Configuration Profile</description>
</property>
<property>
<name>datasourceIdWhitelist</name>
@ -427,8 +419,6 @@
--conf spark.sql.shuffle.partitions=3840
</spark-opts>
<arg>--workingPath</arg><arg>${workingPath}</arg>
<arg>--isLookupUrl</arg><arg>${isLookupUrl}</arg>
<arg>--dedupConfProfile</arg><arg>${dedupConfProfId}</arg>
<arg>--datasourceIdWhitelist</arg><arg>${datasourceIdWhitelist}</arg>
<arg>--datasourceTypeWhitelist</arg><arg>${datasourceTypeWhitelist}</arg>
<arg>--datasourceIdBlacklist</arg><arg>${datasourceIdBlacklist}</arg>

View File

@ -5,18 +5,6 @@
"paramDescription": "the path where the generated events will be stored",
"paramRequired": true
},
{
"paramName": "lu",
"paramLongName": "isLookupUrl",
"paramDescription": "the address of the ISLookUpService",
"paramRequired": true
},
{
"paramName": "d",
"paramLongName": "dedupConfProfile",
"paramDescription": "the id of a valid Dedup Configuration Profile",
"paramRequired": true
},
{
"paramName": "datasourceIdWhitelist",
"paramLongName": "datasourceIdWhitelist",

View File

@ -9,15 +9,6 @@
<name>workingPath</name>
<description>the path where the the generated data will be stored</description>
</property>
<property>
<name>isLookupUrl</name>
<description>the address of the lookUp service</description>
</property>
<property>
<name>dedupConfProfId</name>
<description>the id of a valid Dedup Configuration Profile</description>
</property>
<property>
<name>sparkDriverMemory</name>
<description>memory for driver process</description>

View File

@ -30,7 +30,7 @@ class UpdateMatcherTest {
final OaBrokerMainEntity p4 = new OaBrokerMainEntity();
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.isEmpty());
}
@ -46,7 +46,7 @@ class UpdateMatcherTest {
res.setPublicationdate("2018");
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.isEmpty());
}
@ -62,7 +62,7 @@ class UpdateMatcherTest {
p2.setPublicationdate("2018");
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.size() == 1);
}
@ -79,7 +79,7 @@ class UpdateMatcherTest {
p2.setPublicationdate("2018");
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.isEmpty());
}
@ -98,7 +98,7 @@ class UpdateMatcherTest {
p4.setPublicationdate("2018");
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.isEmpty());
}
@ -117,7 +117,7 @@ class UpdateMatcherTest {
p4.setPublicationdate("2018");
final Collection<UpdateInfo<String>> list = matcher
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null, null);
.searchUpdatesForRecord(res, Arrays.asList(p1, p2, p3, p4), null);
assertTrue(list.size() == 1);
}

View File

@ -5,6 +5,10 @@ import static org.junit.jupiter.api.Assertions.assertTrue;
import org.junit.jupiter.api.Test;
import eu.dnetlib.broker.objects.OaBrokerAuthor;
import eu.dnetlib.broker.objects.OaBrokerMainEntity;
import eu.dnetlib.broker.objects.OaBrokerTypedValue;
public class TrustUtilsTest {
private static final double THRESHOLD = 0.95;
@ -64,6 +68,23 @@ public class TrustUtilsTest {
verifyValue(2.00, BrokerConstants.MAX_TRUST);
}
@Test
public void test() throws Exception {
final OaBrokerMainEntity r1 = new OaBrokerMainEntity();
r1.getTitles().add("D-NET Service Package: Data Import");
r1.getPids().add(new OaBrokerTypedValue("doi", "123"));
r1.getCreators().add(new OaBrokerAuthor("Michele Artini", null));
r1.getCreators().add(new OaBrokerAuthor("Claudio Atzori", null));
final OaBrokerMainEntity r2 = new OaBrokerMainEntity();
r2.getTitles().add("D-NET Service Package: Data Import");
// r2.getPids().add(new OaBrokerTypedValue("doi", "123"));
r2.getCreators().add(new OaBrokerAuthor("Michele Artini", null));
// r2.getCreators().add(new OaBrokerAuthor("Claudio Atzori", null));
System.out.println("TRUST: " + TrustUtils.calculateTrust(r1, r2));
}
private void verifyValue(final double originalScore, final float expectedTrust) {
final float trust = TrustUtils.rescale(originalScore, THRESHOLD);
System.out.println(trust);