1
|
<RESOURCE_PROFILE>
|
2
|
<HEADER>
|
3
|
<RESOURCE_IDENTIFIER value="5f52f22e-b077-43ac-bf22-83de1543c9e1_RGVkdXBDb25maWd1cmF0aW9uRFNSZXNvdXJjZXMvRGVkdXBDb25maWd1cmF0aW9uRFNSZXNvdXJjZVR5cGU="/>
|
4
|
<RESOURCE_TYPE value="DedupConfigurationDSResourceType"/>
|
5
|
<RESOURCE_KIND value="DedupConfigurationDSResources"/>
|
6
|
<RESOURCE_URI value=""/>
|
7
|
<DATE_OF_CREATION value="2001-12-31T12:00:00"/>
|
8
|
</HEADER>
|
9
|
<BODY>
|
10
|
<CONFIGURATION>
|
11
|
<DEDUPLICATION>
|
12
|
<ENTITY name="result">
|
13
|
<PACE>pace.conf {
|
14
|
clustering {
|
15
|
ngrampairs { fields = [title], params = { max = 1, ngramLen = 3} },
|
16
|
suffixprefix { fields = [title], params = { max = 1, len = 3 } } },
|
17
|
conditions {
|
18
|
yearMatch { fields = [dateofacceptance] },
|
19
|
titleVersionMatch { fields = [title] },
|
20
|
sizeMatch { fields = [authors] }
|
21
|
},
|
22
|
model {
|
23
|
title { algo = JaroWinkler, type = String, weight = 1.0, ignoreMissing = false, path = result/metadata/title/value },
|
24
|
dateofacceptance/value { algo = Null, type = String, weight = 0.0, ignoreMissing = true, path = result/metadata/dateofacceptance/value },
|
25
|
authors { algo = Null, type = List, weight = 0.0, ignoreMissing = true, path = result/author/metadata/fullname/value } } }</PACE>
|
26
|
<WORKFLOW>dedup.conf {
|
27
|
threshold = 0.99,
|
28
|
run = '001',
|
29
|
entity.type = result,
|
30
|
order.field = title,
|
31
|
queue.max.size = 2000,
|
32
|
group.max.size = 10,
|
33
|
sliding.window.size = 200,
|
34
|
rootbuilder = [result] }</WORKFLOW>
|
35
|
</ENTITY>
|
36
|
<ENTITY name="person">
|
37
|
<PACE>pace.conf {
|
38
|
clustering { \
|
39
|
ngrampairs { fields = [fullname], params = { max = 1, ngramLen = 3} },
|
40
|
suffixprefix { fields = [fullname], params = { max = 1, len = 3 } } },
|
41
|
model { \
|
42
|
fullname/value { algo = JaroWinkler, type = String, weight = 0.6, ignoreMissing = false },
|
43
|
coauthors/value { algo = JaroWinkler, type = String, weight = 0.4, ignoreMissing = true } } }</PACE>
|
44
|
<WORKFLOW>dedup.conf {
|
45
|
threshold = 0.99,
|
46
|
run = '001',
|
47
|
entity.type = person,
|
48
|
queue.max.size = 2000,
|
49
|
group.max.size = 10,
|
50
|
sliding.window.size = 200,
|
51
|
order.field = fullname, rootbuilder = [person,personResult_authorship_isAuthorOf,projectPerson_contactPerson_isContact] }</WORKFLOW>
|
52
|
</ENTITY>
|
53
|
<ENTITY name="organization">
|
54
|
<PACE>pace.conf {
|
55
|
clustering {
|
56
|
ngrampairs { fields = [legalname], params = { max = 1, ngramLen = 3} },
|
57
|
suffixprefix { fields = [legalname], params = { max = 1, len = 3 } } },
|
58
|
model {
|
59
|
legalname/value { algo = JaroWinkler, type = String, weight = 0.6, ignoreMissing = false },
|
60
|
legalshortname/value { algo = JaroWinkler, type = String, weight = 0.4, ignoreMissing = true } } }</PACE>
|
61
|
<WORKFLOW>dedup.conf {
|
62
|
threshold = 0.99,
|
63
|
run = '001',
|
64
|
entity.type = organization,
|
65
|
order.field = legalname,
|
66
|
queue.max.size = 2000,
|
67
|
group.max.size = 10,
|
68
|
sliding.window.size = 200,
|
69
|
rootbuilder = [organization,projectOrganization_participation_isParticipant,datasourceOrganization_provision_isProvidedBy] }</WORKFLOW>
|
70
|
</ENTITY>
|
71
|
</DEDUPLICATION>
|
72
|
</CONFIGURATION>
|
73
|
<STATUS>
|
74
|
<LAST_UPDATE value="2001-12-31T12:00:00"/>
|
75
|
</STATUS>
|
76
|
<SECURITY_PARAMETERS>SECURITY_PARAMETERS</SECURITY_PARAMETERS>
|
77
|
</BODY>
|
78
|
</RESOURCE_PROFILE>
|