dwc_agent 0.1.0 → 0.1.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/lib/dwc_agent.rb +6 -295
- data/lib/dwc_agent/cleaner.rb +92 -0
- data/lib/dwc_agent/constants.rb +174 -0
- data/lib/dwc_agent/parser.rb +20 -0
- data/lib/dwc_agent/similarity.rb +37 -0
- data/lib/dwc_agent/utility.rb +17 -0
- data/lib/dwc_agent/version.rb +11 -4
- metadata +6 -1
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA1:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: 85a83233676096ae89385fa15d9df82a8ae4fdbb
|
|
4
|
+
data.tar.gz: 5fba52783d9724a2c3d9b6e0011affb204c1e759
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: 1d2a61e2b1d95573b8dc456fb53617f518c07b327fce7773effeba4aef9c8322468e6f8e6cbd72f6107d7f0d59c93b0624b44f54522555542f69cb9ee17a0abb
|
|
7
|
+
data.tar.gz: 8acc297375824ed643db7a1d2ec6b51636d80ee194e2decc8abfc54e3db8d39b4431afb12a25039b5dc27be624676dc4cbdde3fada9c2b6a9907fde0eaeb4ec6
|
data/lib/dwc_agent.rb
CHANGED
|
@@ -1,298 +1,9 @@
|
|
|
1
|
-
require "dwc_agent/version"
|
|
2
1
|
require "capitalize_names"
|
|
3
2
|
require "namae"
|
|
4
3
|
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
|
|
8
|
-
|
|
9
|
-
|
|
10
|
-
|
|
11
|
-
\b[,;]?\s*(?i:and|&)?\s*(?i:others)\s*\b|
|
|
12
|
-
\b[,;]?\s*(?i:etc)\.?|
|
|
13
|
-
\b[,;]?\s*(?i:on)\b|
|
|
14
|
-
\b[,;]?\s*(?i:unkn?own)\b|
|
|
15
|
-
\b[,;]?\s*(?i:n/a)\b|
|
|
16
|
-
\b[,;]?\s*(?i:ann?onymous)\b|
|
|
17
|
-
\b[,;]?\s*(?i:undetermined|indeterminable|dummy|interim)\b|
|
|
18
|
-
\b[,;]?\s*(?i:importer)\b|
|
|
19
|
-
\b[,;]?\s*(?i:frère|frere|père|pere|soeur|sister|bro)\.?(\b|\z)|
|
|
20
|
-
(?i:no\s+(data|disponible))|
|
|
21
|
-
\b[,;]?\s*(?i:stet)[,!]?\s*\d*\z|
|
|
22
|
-
[,;]?\s*\d+[-/\s+](?i:\d+|Jan|Feb|Mar|Apr|
|
|
23
|
-
May|Jun|Jul|Aug|Sept?|
|
|
24
|
-
Oct|Nov|Dec)\.?\s*[-/\s+]?\d+|
|
|
25
|
-
\b[,;]?\s*(?i:Jan|Jan(uary|vier))[.,;]?\s*\d+|
|
|
26
|
-
\b[,;]?\s*(?i:Feb|February|f(é|e)vrier)[.,;]?\s*\d+|
|
|
27
|
-
\b[,;]?\s*(?i:Mar|Mar(ch|s))[.,;]?\s*\d+|
|
|
28
|
-
\b[,;]?\s*(?i:Apr|Apri|April|avril)[.,;]?\s*\d+|
|
|
29
|
-
\b[,;]?\s*(?i:Ma(y|i))[.,;]?\s*\d+|
|
|
30
|
-
\b[,;]?\s*(?i:Jun|June|juin)[.,;]?\s*\d+|
|
|
31
|
-
\b[,;]?\s*(?i:Jul|July|juillet)[.,;]?\s*\d+|
|
|
32
|
-
\b[,;]?\s*(?i:Aug|August|ao(û|u)t)[.,;]?\s*\d+|
|
|
33
|
-
\b[,;]?\s*(?i:Sep|Sept|Septemb(er|re))[.,;]?\s*\d+|
|
|
34
|
-
\b[,;]?\s*(?i:Oct|Octob(er|re))[.,;]?\s*\d+|
|
|
35
|
-
\b[,;]?\s*(?i:Nov|Novemb(er|re))[.,;]?\s*\d+|
|
|
36
|
-
\b[,;]?\s*(?i:Dec|D(é|e)cemb(er|re))[.,;]?\s*\d+|
|
|
37
|
-
\d+\s+(?i:Jan|Jan(uary|vier))\.?\b|
|
|
38
|
-
\d+\s+(?i:Feb|February|f(é|e)vrier)\.?\b|
|
|
39
|
-
\d+\s+(?i:Mar|March|mars)\.?\b|
|
|
40
|
-
\d+\s+(?i:Apr|Apri|April|avril)\.?\b|
|
|
41
|
-
\d+\s+(?i:Ma(y|i))\b|
|
|
42
|
-
\d+\s+(?i:Jun|June|juin)\.?\b|
|
|
43
|
-
\d+\s+(?i:Jul|July|juillet)\.?\b|
|
|
44
|
-
\d+\s+(?i:Aug|August|ao(û|u)t)\.?\b|
|
|
45
|
-
\d+\s+(?i:Sep|Septemb(er|re))t?\.?\b|
|
|
46
|
-
\d+\s+(?i:Oct|Octob(er|re))\.?\b|
|
|
47
|
-
\d+\s+(?i:Nov|Novemb(er|re))\.?\b|
|
|
48
|
-
\d+\s+(?i:Dec|D(e|é)cemb(er|re))\.?\b|
|
|
49
|
-
(?i:autres?\s+de|probab|likely|possibl(e|y)|doubtful)|
|
|
50
|
-
\b\s*maybe\s*\b|
|
|
51
|
-
\(?(?i:collector|data\s*recorder|netter|(oper|prepar)ator)\(?s?\)?\.?\:?|
|
|
52
|
-
(?i:fide)\:?\s*\b|
|
|
53
|
-
(?i:game\s+dept)\.?\s*\b|
|
|
54
|
-
(?i:see\s+notes?\s*(inside)?)|
|
|
55
|
-
(?i:see\s+letter\s+enclosed)|
|
|
56
|
-
(?i:(by)?\s+correspondance)|
|
|
57
|
-
(?i:pers\.?\s+comm\.?)|
|
|
58
|
-
(?i:crossed\s+out)|
|
|
59
|
-
\(?(?i:source)\(?|
|
|
60
|
-
(?i:according\s+to)|
|
|
61
|
-
(?i:revised|photograph|fruits\s+only)|
|
|
62
|
-
-?\s*(?i:sight\s+(id|identifi?cation))\.?\s*\b|
|
|
63
|
-
-?\s*(?i:synonym(y|ie))|
|
|
64
|
-
\b\s*\(?(?i:(fe)?male)\)?\s*\b|
|
|
65
|
-
\b(?i:to\s+(sub)?spp?)\.?|
|
|
66
|
-
(?i:nom\.?\s+rev\.?)|
|
|
67
|
-
FNA|DAO|HUH|\(MT\)|(?i:\(KEW\))|
|
|
68
|
-
(?i:uqam)|
|
|
69
|
-
\b[,;]\s+\d+\z|
|
|
70
|
-
[":!]|
|
|
71
|
-
[,]?\d+|
|
|
72
|
-
\s+\d+?(\/|\.)?(?i:i|ii|iii|iv|v|vi|vii|viii|ix|x)(\/|\.)\d+|
|
|
73
|
-
[,;]\z|
|
|
74
|
-
^\w{0,2}\z|
|
|
75
|
-
^[A-Z]{2,}\z|
|
|
76
|
-
\s+(?i:stet)\s*!?\s*\z|
|
|
77
|
-
\s+(?i:prep)\.?\s*\z|
|
|
78
|
-
\b\s*\([A-Z]{2,}\)
|
|
79
|
-
}x
|
|
80
|
-
|
|
81
|
-
SPLIT_BY = %r{
|
|
82
|
-
[–|&+/;]|
|
|
83
|
-
\s+-\s+|
|
|
84
|
-
\s+a\.\s+|
|
|
85
|
-
\b(?i:and|et|with|per)\s+|
|
|
86
|
-
\be\s*\b|
|
|
87
|
-
\b(?i:annotated(\s+by)?)\s*\b|
|
|
88
|
-
\b(?i:coll\.)\s*\b|
|
|
89
|
-
\b(?i:communicate?d(\s+to)?)\s*\b|
|
|
90
|
-
\b(?i:conf\.?(\s+by)?|confirmed(\s+by)?)\s*\b|
|
|
91
|
-
\b(?i:checked?(\s+by)?)\s*\b|
|
|
92
|
-
\b(?i:det\.?(\s+by)?)\s*\b|
|
|
93
|
-
\b(?i:dupl?\.?(\s+by)?|duplicate(\s+by)?)\s*\b|
|
|
94
|
-
\b(?i:ex\.?(\s+by)?|examined(\s+by)?)\s*\b|
|
|
95
|
-
\b(?i:in?dentified(\s+by)?)\s*\b|
|
|
96
|
-
\b(?i:in\s+part(\s+by)?)\s*\b|
|
|
97
|
-
\b(?i:or)\s+|
|
|
98
|
-
\b(?i:prep\.?\s+(?i:by)?)\s*\b|
|
|
99
|
-
\b(?i:redet\.?(\s+by?)?)\s*\b|
|
|
100
|
-
\b(?i:reidentified(\s+by)?)\s*\b|
|
|
101
|
-
\b(?i:stet)\s*\b|
|
|
102
|
-
\b(?i:then(\s+by)?)\s+|
|
|
103
|
-
\b(?i:veri?f?\.?\:?(\s+by)?|v(e|é)rifi(e|é)d?(\s+by)?)\s*\b|
|
|
104
|
-
\b(?i:via|from)\s*\b|
|
|
105
|
-
\b(?i:(donated)?\s*by)\s+
|
|
106
|
-
}x
|
|
107
|
-
|
|
108
|
-
CHAR_SUBS = {
|
|
109
|
-
'|' => ' | ',
|
|
110
|
-
'(' => ' ',
|
|
111
|
-
')' => ' ',
|
|
112
|
-
'[' => ' ',
|
|
113
|
-
']' => ' ',
|
|
114
|
-
'?' => '',
|
|
115
|
-
'!' => '',
|
|
116
|
-
'=' => '',
|
|
117
|
-
'#' => '',
|
|
118
|
-
'/' => ' / ',
|
|
119
|
-
'&' => ' & ',
|
|
120
|
-
'*' => ''
|
|
121
|
-
}
|
|
122
|
-
|
|
123
|
-
COMPLEX_SEPARATORS = %r{
|
|
124
|
-
^([A-Za-z]{4,},\s+(?:[A-Z]\.\s*){1,})\s+([A-Za-z]{4,},\s+(?:[A-Z]\.\s*){1,})$
|
|
125
|
-
}x
|
|
126
|
-
|
|
127
|
-
BLACKLIST = %r{
|
|
128
|
-
(?i:abundant)|
|
|
129
|
-
(?i:adult|juvenile)|
|
|
130
|
-
(?i:anon)|
|
|
131
|
-
(?i:average)|
|
|
132
|
-
(?i:believe|unclear|illegible|none|suggested|(dis)?agrees?)|approach|
|
|
133
|
-
(?i:barcod)|
|
|
134
|
-
(?i:biolog|botan|zoo|ecolog|mycol|(in)?vertebrate|fisheries|genetic|animal|mushroom|wildlife|plumage|flower|agriculture)|
|
|
135
|
-
(?i:bris?tish|canadi?an?|chinese|arctic|japan|russian|north\s+america)|
|
|
136
|
-
(?i:carex|salix)|
|
|
137
|
-
(?i:herbarium|herbier|collection|collected|publication|specimen|species|describe|an(a|o)morph|isolated|recorded|inspection|define|status|lighthouse)|
|
|
138
|
-
\b\s*(?i:help)\s*\b|
|
|
139
|
-
(?i:description|drawing|identification|remark|original|illustration|checklist|intermedia|measurement|indisting|series|imperfect)|
|
|
140
|
-
(?i:desconocido)|
|
|
141
|
-
(?i:evidence)|
|
|
142
|
-
(?i:inconn?u)|
|
|
143
|
-
(?i:internation|gou?vern|ministry|unit|district|provincial|na(c|t)ional|military|region|environ|natur(e|al)|naturelles|division|program|direction|national)|
|
|
144
|
-
(?i:label)|
|
|
145
|
-
(?i:o?\.?m\.?n\.?r\.?)|
|
|
146
|
-
(?i:measurement)|
|
|
147
|
-
(?i:ent(o|y)mology)|
|
|
148
|
-
(?i:mus(eum|ée)|universit(y|é|e|at)|college|institute?|acad(e|é)m|school|écol(e|iers?)|laboratoi?r|projec?t|polytech|dep(t|art?ment)|research|clinic|hospital|cientifica|sanctuary|safari)|
|
|
149
|
-
(?i:univ\.)|
|
|
150
|
-
(?i:graduate|student|estudantes|storekeep|supervisor|superint|rcmp|coordinator|minority|fishermen|police|taxonomist|consultant|participante?s?|team|(é|e)quipe|memb(er|re)|crew|group|staff|personnel|family|captain|friends|assistant|worker)|
|
|
151
|
-
(?i:non\s+pr(é|e)cis(é|e))|
|
|
152
|
-
(?i:ontario|qu(e|é)bec|saskatchewan|new brunswick|sault|newfoundland|assurance|vancouver|u\.?s\.?s\.?r\.?)|
|
|
153
|
-
(?i:recreation|culture)|
|
|
154
|
-
(?i:shaped|dark|pale|areas|phase|spotting|interior|between|closer)|
|
|
155
|
-
(?i:soci(e|é)t(y|é)|cent(er|re)|community|history|conservation|conference|assoc|class|commission|consortium|council|club|alliance|protective|circle)|
|
|
156
|
-
(?i:commercial|company|control|product)|
|
|
157
|
-
(?i:size|large|colou?r)\s+|
|
|
158
|
-
(?i:skeleton)|
|
|
159
|
-
(?i:survey|assessment|station|monitor|stn\.|index|project|bureau|engine|expedi(c|t)ion|festival|generation|inventory|marine|service)|
|
|
160
|
-
(?i:submersible)|
|
|
161
|
-
(?i:synonymy?)|(topo|syn|holo)type|
|
|
162
|
-
(?i:systematic|perspective)|
|
|
163
|
-
\s+(?i:off)\s+|
|
|
164
|
-
\s*(?i:too)\s+|\s*(?i:the)\s+|
|
|
165
|
-
(?i:taxiderm(ies|y))|
|
|
166
|
-
(?i:though)|
|
|
167
|
-
(?i:toward|seen at)|
|
|
168
|
-
(?i:unidentified|unspecified|unk?nown|unnamed|unread|unmistak|no agent)|
|
|
169
|
-
(?i:urn\:)|
|
|
170
|
-
(?i:usda|ucla)|
|
|
171
|
-
(?i:workshop|garden|farm|jardin|public)
|
|
172
|
-
}x
|
|
173
|
-
|
|
174
|
-
TITLE = /\s*\b(sir|lord|count(ess)?|(gen|adm|col|maj|capt|cmdr|lt|sgt|cpl|pvt|prof|dr|md|ph\.?d|rev|docteur|mme|abbé|ptre)\.?|frère|frere|père|pere|professor)(\s+|$)/i
|
|
175
|
-
|
|
176
|
-
Namae.options[:prefer_comma_as_separator] = true
|
|
177
|
-
Namae.options[:separator] = SPLIT_BY
|
|
178
|
-
Namae.options[:title] = TITLE
|
|
179
|
-
|
|
180
|
-
def self.parse(name)
|
|
181
|
-
return [] if name.nil? || name == ""
|
|
182
|
-
cleaned = name.gsub(STRIP_OUT, ' ')
|
|
183
|
-
.gsub(/[#{CHAR_SUBS.keys.join('\\')}]/, CHAR_SUBS)
|
|
184
|
-
.gsub(/([A-Z]{1}\.)([[:alpha:]]{2,})/, '\1 \2')
|
|
185
|
-
.gsub(COMPLEX_SEPARATORS, '\1 | \2')
|
|
186
|
-
.gsub(/,\z/, '')
|
|
187
|
-
.squeeze(' ').strip
|
|
188
|
-
Namae.parse(cleaned)
|
|
189
|
-
end
|
|
190
|
-
|
|
191
|
-
def self.clean(parsed_namae)
|
|
192
|
-
blank_name = { given: nil, family: nil }
|
|
193
|
-
|
|
194
|
-
if parsed_namae.family && parsed_namae.family.length < 3
|
|
195
|
-
return blank_name
|
|
196
|
-
end
|
|
197
|
-
if parsed_namae.family && parsed_namae.family.length == 3 && parsed_namae.family.count('.') == 1
|
|
198
|
-
return blank_name
|
|
199
|
-
end
|
|
200
|
-
if parsed_namae.given && parsed_namae.given.length > 15
|
|
201
|
-
return blank_name
|
|
202
|
-
end
|
|
203
|
-
if parsed_namae.given && parsed_namae.given.count('.') >= 3 && /\.\s*[a-zA-Z]{4,}\s+[a-zA-Z]{1,}\./.match(parsed_namae.given)
|
|
204
|
-
return blank_name
|
|
205
|
-
end
|
|
206
|
-
if parsed_namae.family && /[a-zA-Z]{2,}\.?\s+[a-zA-Z]{2,}/.match(parsed_namae.family)
|
|
207
|
-
return blank_name
|
|
208
|
-
end
|
|
209
|
-
if parsed_namae.given && /[a-zA-Z]{2,}\.?\s+[a-zA-Z]{2,}/.match(parsed_namae.given)
|
|
210
|
-
return blank_name
|
|
211
|
-
end
|
|
212
|
-
if parsed_namae.display_order =~ BLACKLIST
|
|
213
|
-
return blank_name
|
|
214
|
-
end
|
|
215
|
-
|
|
216
|
-
if parsed_namae.given &&
|
|
217
|
-
parsed_namae.family &&
|
|
218
|
-
parsed_namae.family.count(".") > 0 &&
|
|
219
|
-
parsed_namae.family.length - parsed_namae.family.count(".") <= 3
|
|
220
|
-
given = parsed_namae.given
|
|
221
|
-
family = parsed_namae.family
|
|
222
|
-
parsed_namae.family = given
|
|
223
|
-
parsed_namae.given = family
|
|
224
|
-
end
|
|
225
|
-
|
|
226
|
-
if parsed_namae.given &&
|
|
227
|
-
(parsed_namae.given == parsed_namae.given.upcase ||
|
|
228
|
-
parsed_namae.given == parsed_namae.given.downcase) &&
|
|
229
|
-
!parsed_namae.given.include?(".") &&
|
|
230
|
-
parsed_namae.given.length >= 4
|
|
231
|
-
parsed_namae.given = CapitalizeNames.capitalize(parsed_namae.given)
|
|
232
|
-
end
|
|
233
|
-
|
|
234
|
-
if parsed_namae.given && /[A-Za-z]\./.match(parsed_namae.given)
|
|
235
|
-
parsed_namae.given = CapitalizeNames.capitalize(parsed_namae.given).sub(/[a-z]\./, &:upcase)
|
|
236
|
-
end
|
|
237
|
-
|
|
238
|
-
parsed_namae.normalize_initials
|
|
239
|
-
|
|
240
|
-
family = parsed_namae.family.gsub(/\.\z/, '').strip rescue nil
|
|
241
|
-
given = parsed_namae.given.strip rescue nil
|
|
242
|
-
particle = parsed_namae.particle.strip rescue nil
|
|
243
|
-
|
|
244
|
-
if family.nil? && !given.nil? && !given.include?(".")
|
|
245
|
-
family = given
|
|
246
|
-
given = nil
|
|
247
|
-
end
|
|
248
|
-
|
|
249
|
-
if !family.nil? && given.nil? && !particle.nil?
|
|
250
|
-
given = particle.sub(/[a-z]\./, &:upcase).sub(/^(.)/) { $1.capitalize }
|
|
251
|
-
particle = nil
|
|
252
|
-
end
|
|
253
|
-
|
|
254
|
-
if !family.nil? && (family == family.upcase || family == family.downcase)
|
|
255
|
-
family = family.mb_chars.capitalize.to_s rescue nil
|
|
256
|
-
end
|
|
257
|
-
|
|
258
|
-
if !family.nil? && family.length <= 4 && family !~ /[aeiouy]/
|
|
259
|
-
return blank_name
|
|
260
|
-
end
|
|
261
|
-
|
|
262
|
-
if !family.nil? && family.match(/[A-Z]$/)
|
|
263
|
-
return blank_name
|
|
264
|
-
end
|
|
265
|
-
|
|
266
|
-
if !family.nil? && family.match(/^[A-Z]{2}/)
|
|
267
|
-
return blank_name
|
|
268
|
-
end
|
|
269
|
-
|
|
270
|
-
{ given: given, family: family }
|
|
271
|
-
end
|
|
272
|
-
|
|
273
|
-
def self.similarity_score(given1, given2)
|
|
274
|
-
given1_parts = given1.gsub(/\.\s+/,".").split(/[\.\s]/)
|
|
275
|
-
given2_parts = given2.gsub(/\.\s+/,".").split(/[\.\s]/)
|
|
276
|
-
largest = [given1_parts,given2_parts].max
|
|
277
|
-
smallest = [given1_parts,given2_parts].min
|
|
278
|
-
|
|
279
|
-
score = 0
|
|
280
|
-
largest.each_with_index do |val,index|
|
|
281
|
-
if smallest[index]
|
|
282
|
-
if val[0] == smallest[index][0]
|
|
283
|
-
score += 1
|
|
284
|
-
else
|
|
285
|
-
return 0
|
|
286
|
-
end
|
|
287
|
-
if val.length > 1 && smallest[index].length > 1 && val != smallest[index]
|
|
288
|
-
return 0
|
|
289
|
-
end
|
|
290
|
-
else
|
|
291
|
-
score += 0.1
|
|
292
|
-
end
|
|
293
|
-
end
|
|
294
|
-
|
|
295
|
-
score
|
|
296
|
-
end
|
|
297
|
-
|
|
298
|
-
end
|
|
4
|
+
require "dwc_agent/version"
|
|
5
|
+
require "dwc_agent/constants"
|
|
6
|
+
require "dwc_agent/cleaner"
|
|
7
|
+
require "dwc_agent/parser"
|
|
8
|
+
require "dwc_agent/similarity"
|
|
9
|
+
require "dwc_agent/utility"
|
|
@@ -0,0 +1,92 @@
|
|
|
1
|
+
module DwcAgent
|
|
2
|
+
class Cleaner
|
|
3
|
+
|
|
4
|
+
# Cleans the passed-in namae object from the parse method and
|
|
5
|
+
# re-organizes it to better match expected Darwin Core output.
|
|
6
|
+
#
|
|
7
|
+
# @param parsed_namae [Object] the namae object
|
|
8
|
+
# @return [Hash] the given, family hash
|
|
9
|
+
def self.clean(parsed_namae)
|
|
10
|
+
blank_name = { given: nil, family: nil }
|
|
11
|
+
|
|
12
|
+
if parsed_namae.family && parsed_namae.family.length < 3
|
|
13
|
+
return blank_name
|
|
14
|
+
end
|
|
15
|
+
if parsed_namae.family && parsed_namae.family.length == 3 && parsed_namae.family.count('.') == 1
|
|
16
|
+
return blank_name
|
|
17
|
+
end
|
|
18
|
+
if parsed_namae.given && parsed_namae.given.length > 15
|
|
19
|
+
return blank_name
|
|
20
|
+
end
|
|
21
|
+
if parsed_namae.given && parsed_namae.given.count('.') >= 3 && /\.\s*[a-zA-Z]{4,}\s+[a-zA-Z]{1,}\./.match(parsed_namae.given)
|
|
22
|
+
return blank_name
|
|
23
|
+
end
|
|
24
|
+
if parsed_namae.family && /[a-zA-Z]{2,}\.?\s+[a-zA-Z]{2,}/.match(parsed_namae.family)
|
|
25
|
+
return blank_name
|
|
26
|
+
end
|
|
27
|
+
if parsed_namae.given && /[a-zA-Z]{2,}\.?\s+[a-zA-Z]{2,}/.match(parsed_namae.given)
|
|
28
|
+
return blank_name
|
|
29
|
+
end
|
|
30
|
+
if parsed_namae.display_order =~ BLACKLIST
|
|
31
|
+
return blank_name
|
|
32
|
+
end
|
|
33
|
+
|
|
34
|
+
if parsed_namae.given &&
|
|
35
|
+
parsed_namae.family &&
|
|
36
|
+
parsed_namae.family.count(".") > 0 &&
|
|
37
|
+
parsed_namae.family.length - parsed_namae.family.count(".") <= 3
|
|
38
|
+
given = parsed_namae.given
|
|
39
|
+
family = parsed_namae.family
|
|
40
|
+
parsed_namae.family = given
|
|
41
|
+
parsed_namae.given = family
|
|
42
|
+
end
|
|
43
|
+
|
|
44
|
+
if parsed_namae.given &&
|
|
45
|
+
(parsed_namae.given == parsed_namae.given.upcase ||
|
|
46
|
+
parsed_namae.given == parsed_namae.given.downcase) &&
|
|
47
|
+
!parsed_namae.given.include?(".") &&
|
|
48
|
+
parsed_namae.given.length >= 4
|
|
49
|
+
parsed_namae.given = CapitalizeNames.capitalize(parsed_namae.given)
|
|
50
|
+
end
|
|
51
|
+
|
|
52
|
+
if parsed_namae.given && /[A-Za-z]\./.match(parsed_namae.given)
|
|
53
|
+
parsed_namae.given = CapitalizeNames.capitalize(parsed_namae.given).sub(/[a-z]\./, &:upcase)
|
|
54
|
+
end
|
|
55
|
+
|
|
56
|
+
parsed_namae.normalize_initials
|
|
57
|
+
|
|
58
|
+
family = parsed_namae.family.gsub(/\.\z/, '').strip rescue nil
|
|
59
|
+
given = parsed_namae.given.strip rescue nil
|
|
60
|
+
particle = parsed_namae.particle.strip rescue nil
|
|
61
|
+
|
|
62
|
+
if family.nil? && !given.nil? && !given.include?(".")
|
|
63
|
+
family = given
|
|
64
|
+
given = nil
|
|
65
|
+
end
|
|
66
|
+
|
|
67
|
+
if !family.nil? && given.nil? && !particle.nil?
|
|
68
|
+
given = particle.sub(/[a-z]\./, &:upcase).sub(/^(.)/) { $1.capitalize }
|
|
69
|
+
particle = nil
|
|
70
|
+
end
|
|
71
|
+
|
|
72
|
+
if !family.nil? && (family == family.upcase || family == family.downcase)
|
|
73
|
+
family = family.mb_chars.capitalize.to_s rescue nil
|
|
74
|
+
end
|
|
75
|
+
|
|
76
|
+
if !family.nil? && family.length <= 4 && family !~ /[aeiouy]/
|
|
77
|
+
return blank_name
|
|
78
|
+
end
|
|
79
|
+
|
|
80
|
+
if !family.nil? && family.match(/[A-Z]$/)
|
|
81
|
+
return blank_name
|
|
82
|
+
end
|
|
83
|
+
|
|
84
|
+
if !family.nil? && family.match(/^[A-Z]{2}/)
|
|
85
|
+
return blank_name
|
|
86
|
+
end
|
|
87
|
+
|
|
88
|
+
{ given: given, family: family }
|
|
89
|
+
end
|
|
90
|
+
|
|
91
|
+
end
|
|
92
|
+
end
|
|
@@ -0,0 +1,174 @@
|
|
|
1
|
+
module DwcAgent
|
|
2
|
+
STRIP_OUT = %r{
|
|
3
|
+
\b\d+\(?(?i:[[:alpha:]])\)?\b|
|
|
4
|
+
\b[,;]?\s*(?i:et\s+al)\.?|
|
|
5
|
+
\bu\.\s*a\.|
|
|
6
|
+
\b[,;]?\s*(?i:and|&)?\s*(?i:others)\s*\b|
|
|
7
|
+
\b[,;]?\s*(?i:etc)\.?|
|
|
8
|
+
\b[,;]?\s*(?i:on)\b|
|
|
9
|
+
\b[,;]?\s*(?i:unkn?own)\b|
|
|
10
|
+
\b[,;]?\s*(?i:n/a)\b|
|
|
11
|
+
\b[,;]?\s*(?i:ann?onymous)\b|
|
|
12
|
+
\b[,;]?\s*(?i:undetermined|indeterminable|dummy|interim)\b|
|
|
13
|
+
\b[,;]?\s*(?i:importer)\b|
|
|
14
|
+
\b[,;]?\s*(?i:frère|frere|père|pere|soeur|sister|bro)\.?(\b|\z)|
|
|
15
|
+
(?i:no\s+(data|disponible))|
|
|
16
|
+
\b[,;]?\s*(?i:stet)[,!]?\s*\d*\z|
|
|
17
|
+
[,;]?\s*\d+[-/\s+](?i:\d+|Jan|Feb|Mar|Apr|
|
|
18
|
+
May|Jun|Jul|Aug|Sept?|
|
|
19
|
+
Oct|Nov|Dec)\.?\s*[-/\s+]?\d+|
|
|
20
|
+
\b[,;]?\s*(?i:Jan|Jan(uary|vier))[.,;]?\s*\d+|
|
|
21
|
+
\b[,;]?\s*(?i:Feb|February|f(é|e)vrier)[.,;]?\s*\d+|
|
|
22
|
+
\b[,;]?\s*(?i:Mar|Mar(ch|s))[.,;]?\s*\d+|
|
|
23
|
+
\b[,;]?\s*(?i:Apr|Apri|April|avril)[.,;]?\s*\d+|
|
|
24
|
+
\b[,;]?\s*(?i:Ma(y|i))[.,;]?\s*\d+|
|
|
25
|
+
\b[,;]?\s*(?i:Jun|June|juin)[.,;]?\s*\d+|
|
|
26
|
+
\b[,;]?\s*(?i:Jul|July|juillet)[.,;]?\s*\d+|
|
|
27
|
+
\b[,;]?\s*(?i:Aug|August|ao(û|u)t)[.,;]?\s*\d+|
|
|
28
|
+
\b[,;]?\s*(?i:Sep|Sept|Septemb(er|re))[.,;]?\s*\d+|
|
|
29
|
+
\b[,;]?\s*(?i:Oct|Octob(er|re))[.,;]?\s*\d+|
|
|
30
|
+
\b[,;]?\s*(?i:Nov|Novemb(er|re))[.,;]?\s*\d+|
|
|
31
|
+
\b[,;]?\s*(?i:Dec|D(é|e)cemb(er|re))[.,;]?\s*\d+|
|
|
32
|
+
\d+\s+(?i:Jan|Jan(uary|vier))\.?\b|
|
|
33
|
+
\d+\s+(?i:Feb|February|f(é|e)vrier)\.?\b|
|
|
34
|
+
\d+\s+(?i:Mar|March|mars)\.?\b|
|
|
35
|
+
\d+\s+(?i:Apr|Apri|April|avril)\.?\b|
|
|
36
|
+
\d+\s+(?i:Ma(y|i))\b|
|
|
37
|
+
\d+\s+(?i:Jun|June|juin)\.?\b|
|
|
38
|
+
\d+\s+(?i:Jul|July|juillet)\.?\b|
|
|
39
|
+
\d+\s+(?i:Aug|August|ao(û|u)t)\.?\b|
|
|
40
|
+
\d+\s+(?i:Sep|Septemb(er|re))t?\.?\b|
|
|
41
|
+
\d+\s+(?i:Oct|Octob(er|re))\.?\b|
|
|
42
|
+
\d+\s+(?i:Nov|Novemb(er|re))\.?\b|
|
|
43
|
+
\d+\s+(?i:Dec|D(e|é)cemb(er|re))\.?\b|
|
|
44
|
+
(?i:autres?\s+de|probab|likely|possibl(e|y)|doubtful)|
|
|
45
|
+
\b\s*maybe\s*\b|
|
|
46
|
+
\(?(?i:collector|data\s*recorder|netter|(oper|prepar)ator)\(?s?\)?\.?\:?|
|
|
47
|
+
(?i:fide)\:?\s*\b|
|
|
48
|
+
(?i:game\s+dept)\.?\s*\b|
|
|
49
|
+
(?i:see\s+notes?\s*(inside)?)|
|
|
50
|
+
(?i:see\s+letter\s+enclosed)|
|
|
51
|
+
(?i:(by)?\s+correspondance)|
|
|
52
|
+
(?i:pers\.?\s+comm\.?)|
|
|
53
|
+
(?i:crossed\s+out)|
|
|
54
|
+
\(?(?i:source)\(?|
|
|
55
|
+
(?i:according\s+to)|
|
|
56
|
+
(?i:revised|photograph|fruits\s+only)|
|
|
57
|
+
-?\s*(?i:sight\s+(id|identifi?cation))\.?\s*\b|
|
|
58
|
+
-?\s*(?i:synonym(y|ie))|
|
|
59
|
+
\b\s*\(?(?i:(fe)?male)\)?\s*\b|
|
|
60
|
+
\b(?i:to\s+(sub)?spp?)\.?|
|
|
61
|
+
(?i:nom\.?\s+rev\.?)|
|
|
62
|
+
FNA|DAO|HUH|\(MT\)|(?i:\(KEW\))|
|
|
63
|
+
(?i:uqam)|
|
|
64
|
+
\b[,;]\s+\d+\z|
|
|
65
|
+
[":!]|
|
|
66
|
+
[,]?\d+|
|
|
67
|
+
\s+\d+?(\/|\.)?(?i:i|ii|iii|iv|v|vi|vii|viii|ix|x)(\/|\.)\d+|
|
|
68
|
+
[,;]\z|
|
|
69
|
+
^\w{0,2}\z|
|
|
70
|
+
^[A-Z]{2,}\z|
|
|
71
|
+
\s+(?i:stet)\s*!?\s*\z|
|
|
72
|
+
\s+(?i:prep)\.?\s*\z|
|
|
73
|
+
\b\s*\([A-Z]{2,}\)
|
|
74
|
+
}x
|
|
75
|
+
|
|
76
|
+
SPLIT_BY = %r{
|
|
77
|
+
[–|&+/;]|
|
|
78
|
+
\s+-\s+|
|
|
79
|
+
\s+a\.\s+|
|
|
80
|
+
\b(?i:and|et|with|per)\s+|
|
|
81
|
+
\be\s*\b|
|
|
82
|
+
\b(?i:annotated(\s+by)?)\s*\b|
|
|
83
|
+
\b(?i:coll\.)\s*\b|
|
|
84
|
+
\b(?i:communicate?d(\s+to)?)\s*\b|
|
|
85
|
+
\b(?i:conf\.?(\s+by)?|confirmed(\s+by)?)\s*\b|
|
|
86
|
+
\b(?i:checked?(\s+by)?)\s*\b|
|
|
87
|
+
\b(?i:det\.?(\s+by)?)\s*\b|
|
|
88
|
+
\b(?i:dupl?\.?(\s+by)?|duplicate(\s+by)?)\s*\b|
|
|
89
|
+
\b(?i:ex\.?(\s+by)?|examined(\s+by)?)\s*\b|
|
|
90
|
+
\b(?i:in?dentified(\s+by)?)\s*\b|
|
|
91
|
+
\b(?i:in\s+part(\s+by)?)\s*\b|
|
|
92
|
+
\b(?i:or)\s+|
|
|
93
|
+
\b(?i:prep\.?\s+(?i:by)?)\s*\b|
|
|
94
|
+
\b(?i:redet\.?(\s+by?)?)\s*\b|
|
|
95
|
+
\b(?i:reidentified(\s+by)?)\s*\b|
|
|
96
|
+
\b(?i:stet)\s*\b|
|
|
97
|
+
\b(?i:then(\s+by)?)\s+|
|
|
98
|
+
\b(?i:veri?f?\.?\:?(\s+by)?|v(e|é)rifi(e|é)d?(\s+by)?)\s*\b|
|
|
99
|
+
\b(?i:via|from)\s*\b|
|
|
100
|
+
\b(?i:(donated)?\s*by)\s+
|
|
101
|
+
}x
|
|
102
|
+
|
|
103
|
+
CHAR_SUBS = {
|
|
104
|
+
'|' => ' | ',
|
|
105
|
+
'(' => ' ',
|
|
106
|
+
')' => ' ',
|
|
107
|
+
'[' => ' ',
|
|
108
|
+
']' => ' ',
|
|
109
|
+
'?' => '',
|
|
110
|
+
'!' => '',
|
|
111
|
+
'=' => '',
|
|
112
|
+
'#' => '',
|
|
113
|
+
'/' => ' / ',
|
|
114
|
+
'&' => ' & ',
|
|
115
|
+
'*' => ''
|
|
116
|
+
}
|
|
117
|
+
|
|
118
|
+
COMPLEX_SEPARATORS = %r{
|
|
119
|
+
^([A-Za-z]{4,},\s+(?:[A-Z]\.\s*){1,})\s+([A-Za-z]{4,},\s+(?:[A-Z]\.\s*){1,})$
|
|
120
|
+
}x
|
|
121
|
+
|
|
122
|
+
BLACKLIST = %r{
|
|
123
|
+
(?i:abundant)|
|
|
124
|
+
(?i:adult|juvenile)|
|
|
125
|
+
(?i:anon)|
|
|
126
|
+
(?i:average)|
|
|
127
|
+
(?i:believe|unclear|illegible|none|suggested|(dis)?agrees?)|approach|
|
|
128
|
+
(?i:barcod)|
|
|
129
|
+
(?i:biolog|botan|zoo|ecolog|mycol|(in)?vertebrate|fisheries|genetic|animal|mushroom|wildlife|plumage|flower|agriculture)|
|
|
130
|
+
(?i:bris?tish|canadi?an?|chinese|arctic|japan|russian|north\s+america)|
|
|
131
|
+
(?i:carex|salix)|
|
|
132
|
+
(?i:herbarium|herbier|collection|collected|publication|specimen|species|describe|an(a|o)morph|isolated|recorded|inspection|define|status|lighthouse)|
|
|
133
|
+
\b\s*(?i:help)\s*\b|
|
|
134
|
+
(?i:description|drawing|identification|remark|original|illustration|checklist|intermedia|measurement|indisting|series|imperfect)|
|
|
135
|
+
(?i:desconocido)|
|
|
136
|
+
(?i:evidence)|
|
|
137
|
+
(?i:inconn?u)|
|
|
138
|
+
(?i:internation|gou?vern|ministry|unit|district|provincial|na(c|t)ional|military|region|environ|natur(e|al)|naturelles|division|program|direction|national)|
|
|
139
|
+
(?i:label)|
|
|
140
|
+
(?i:o?\.?m\.?n\.?r\.?)|
|
|
141
|
+
(?i:measurement)|
|
|
142
|
+
(?i:ent(o|y)mology)|
|
|
143
|
+
(?i:mus(eum|ée)|universit(y|é|e|at)|college|institute?|acad(e|é)m|school|écol(e|iers?)|laboratoi?r|projec?t|polytech|dep(t|art?ment)|research|clinic|hospital|cientifica|sanctuary|safari)|
|
|
144
|
+
(?i:univ\.)|
|
|
145
|
+
(?i:graduate|student|estudantes|storekeep|supervisor|superint|rcmp|coordinator|minority|fishermen|police|taxonomist|consultant|participante?s?|team|(é|e)quipe|memb(er|re)|crew|group|staff|personnel|family|captain|friends|assistant|worker)|
|
|
146
|
+
(?i:non\s+pr(é|e)cis(é|e))|
|
|
147
|
+
(?i:ontario|qu(e|é)bec|saskatchewan|new brunswick|sault|newfoundland|assurance|vancouver|u\.?s\.?s\.?r\.?)|
|
|
148
|
+
(?i:recreation|culture)|
|
|
149
|
+
(?i:shaped|dark|pale|areas|phase|spotting|interior|between|closer)|
|
|
150
|
+
(?i:soci(e|é)t(y|é)|cent(er|re)|community|history|conservation|conference|assoc|class|commission|consortium|council|club|alliance|protective|circle)|
|
|
151
|
+
(?i:commercial|company|control|product)|
|
|
152
|
+
(?i:size|large|colou?r)\s+|
|
|
153
|
+
(?i:skeleton)|
|
|
154
|
+
(?i:survey|assessment|station|monitor|stn\.|index|project|bureau|engine|expedi(c|t)ion|festival|generation|inventory|marine|service)|
|
|
155
|
+
(?i:submersible)|
|
|
156
|
+
(?i:synonymy?)|(topo|syn|holo)type|
|
|
157
|
+
(?i:systematic|perspective)|
|
|
158
|
+
\s+(?i:off)\s+|
|
|
159
|
+
\s*(?i:too)\s+|\s*(?i:the)\s+|
|
|
160
|
+
(?i:taxiderm(ies|y))|
|
|
161
|
+
(?i:though)|
|
|
162
|
+
(?i:toward|seen at)|
|
|
163
|
+
(?i:unidentified|unspecified|unk?nown|unnamed|unread|unmistak|no agent)|
|
|
164
|
+
(?i:urn\:)|
|
|
165
|
+
(?i:usda|ucla)|
|
|
166
|
+
(?i:workshop|garden|farm|jardin|public)
|
|
167
|
+
}x
|
|
168
|
+
|
|
169
|
+
TITLE = /\s*\b(sir|lord|count(ess)?|(gen|adm|col|maj|capt|cmdr|lt|sgt|cpl|pvt|prof|dr|md|ph\.?d|rev|docteur|mme|abbé|ptre)\.?|frère|frere|père|pere|professor)(\s+|$)/i
|
|
170
|
+
|
|
171
|
+
Namae.options[:prefer_comma_as_separator] = true
|
|
172
|
+
Namae.options[:separator] = SPLIT_BY
|
|
173
|
+
Namae.options[:title] = TITLE
|
|
174
|
+
end
|
|
@@ -0,0 +1,20 @@
|
|
|
1
|
+
module DwcAgent
|
|
2
|
+
class Parser
|
|
3
|
+
|
|
4
|
+
# Parses the passed-in string and returns a list of names.
|
|
5
|
+
#
|
|
6
|
+
# @param names [String] the name or names to be parsed
|
|
7
|
+
# @return [Array] the list of parsed names
|
|
8
|
+
def self.parse(name)
|
|
9
|
+
return [] if name.nil? || name == ""
|
|
10
|
+
cleaned = name.gsub(STRIP_OUT, ' ')
|
|
11
|
+
.gsub(/[#{CHAR_SUBS.keys.join('\\')}]/, CHAR_SUBS)
|
|
12
|
+
.gsub(/([A-Z]{1}\.)([[:alpha:]]{2,})/, '\1 \2')
|
|
13
|
+
.gsub(COMPLEX_SEPARATORS, '\1 | \2')
|
|
14
|
+
.gsub(/,\z/, '')
|
|
15
|
+
.squeeze(' ').strip
|
|
16
|
+
Namae.parse(cleaned)
|
|
17
|
+
end
|
|
18
|
+
|
|
19
|
+
end
|
|
20
|
+
end
|
|
@@ -0,0 +1,37 @@
|
|
|
1
|
+
module DwcAgent
|
|
2
|
+
class Similarity
|
|
3
|
+
|
|
4
|
+
# Produces a similarity score of two given names
|
|
5
|
+
# Logic inspired by R.D.M. Page, https://orcid.org/0000-0002-7101-9767
|
|
6
|
+
# At https://linen-baseball.glitch.me/
|
|
7
|
+
#
|
|
8
|
+
# @param given1 [String] one given name
|
|
9
|
+
# @param given2 [String] a second given name
|
|
10
|
+
# @return [Float] the similarity score
|
|
11
|
+
def self.similarity_score(given1, given2)
|
|
12
|
+
given1_parts = given1.gsub(/\.\s+/,".").split(/[\.\s]/)
|
|
13
|
+
given2_parts = given2.gsub(/\.\s+/,".").split(/[\.\s]/)
|
|
14
|
+
largest = [given1_parts,given2_parts].max
|
|
15
|
+
smallest = [given1_parts,given2_parts].min
|
|
16
|
+
|
|
17
|
+
score = 0
|
|
18
|
+
largest.each_with_index do |val,index|
|
|
19
|
+
if smallest[index]
|
|
20
|
+
if val[0] == smallest[index][0]
|
|
21
|
+
score += 1
|
|
22
|
+
else
|
|
23
|
+
return 0
|
|
24
|
+
end
|
|
25
|
+
if val.length > 1 && smallest[index].length > 1 && val != smallest[index]
|
|
26
|
+
return 0
|
|
27
|
+
end
|
|
28
|
+
else
|
|
29
|
+
score += 0.1
|
|
30
|
+
end
|
|
31
|
+
end
|
|
32
|
+
|
|
33
|
+
score
|
|
34
|
+
end
|
|
35
|
+
|
|
36
|
+
end
|
|
37
|
+
end
|
data/lib/dwc_agent/version.rb
CHANGED
|
@@ -1,7 +1,14 @@
|
|
|
1
|
-
|
|
2
|
-
|
|
1
|
+
module DwcAgent
|
|
2
|
+
class Version
|
|
3
|
+
|
|
4
|
+
MAJOR = 0
|
|
5
|
+
MINOR = 1
|
|
6
|
+
PATCH = 1
|
|
7
|
+
BUILD = nil
|
|
8
|
+
|
|
9
|
+
def self.version
|
|
10
|
+
[MAJOR, MINOR, PATCH, BUILD].compact.join('.').freeze
|
|
11
|
+
end
|
|
3
12
|
|
|
4
|
-
def self.version
|
|
5
|
-
VERSION
|
|
6
13
|
end
|
|
7
14
|
end
|
metadata
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
|
2
2
|
name: dwc_agent
|
|
3
3
|
version: !ruby/object:Gem::Version
|
|
4
|
-
version: 0.1.
|
|
4
|
+
version: 0.1.1
|
|
5
5
|
platform: ruby
|
|
6
6
|
authors:
|
|
7
7
|
- David P. Shorthouse
|
|
@@ -90,6 +90,11 @@ extra_rdoc_files: []
|
|
|
90
90
|
files:
|
|
91
91
|
- bin/dwcagent
|
|
92
92
|
- lib/dwc_agent.rb
|
|
93
|
+
- lib/dwc_agent/cleaner.rb
|
|
94
|
+
- lib/dwc_agent/constants.rb
|
|
95
|
+
- lib/dwc_agent/parser.rb
|
|
96
|
+
- lib/dwc_agent/similarity.rb
|
|
97
|
+
- lib/dwc_agent/utility.rb
|
|
93
98
|
- lib/dwc_agent/version.rb
|
|
94
99
|
homepage: https://github.com/dshorthouse/dwc_agent
|
|
95
100
|
licenses:
|