CSV_to_RDF_onomastica_datini_lettere_mani.py 24 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436
  1. #Parser to convert the Datini onomastics CSV file into TTL format
  2. # Utilities to read/write csv files
  3. import csv
  4. # Utilities to handle character encodings
  5. import unicodedata
  6. # Ordered Dicts
  7. from collections import OrderedDict
  8. import json
  9. import re
  10. # OPZIONAL IMPORTS
  11. # For timestamping/simple speed tests
  12. from datetime import datetime
  13. # Random number generator
  14. from random import *
  15. # System & command line utilities
  16. import sys
  17. # Json for the dictionary
  18. import json
  19. import_dir = '/Users/federicaspinelli/TEAMOVI/Parser/DATA/ASPO/CSV/datini/'
  20. export_dir = '/Users/federicaspinelli/TEAMOVI/Parser/DATA/ASPO/RDF/datini/'
  21. # Custom class to store URIs + related infos for the ontologies/repositories
  22. class RDFcoords:
  23. def __init__(self, uri, prefix, code = None):
  24. self.uri = uri
  25. self.prefix = prefix
  26. self.code = code
  27. # Repositories
  28. aspoCoords = RDFcoords('<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/>', 'aspo:')
  29. foafCoords = RDFcoords('<http://xmlns.com/foaf/0.1/>', 'foaf:')
  30. cidocCoords = RDFcoords('<http://www.cidoc-crm.org/cidoc-crm/>', 'crm:')
  31. schemaCoords = RDFcoords('<http://schema.org/>', 'schema:')
  32. personCoords = RDFcoords('<http://www.w3.org/ns/person#>', 'person:')
  33. nsCoords = RDFcoords('<http://www.w3.org/1999/02/22-rdf-syntax-ns#>', 'rdf:')
  34. rdfsCoords = RDFcoords('<http://www.w3.org/2000/01/rdf-schema#>', 'rdfs:')
  35. owlCoords = RDFcoords('<http://www.w3.org/2002/07/owl#>', 'owl:')
  36. # Basic functions for triples / shortened triples in TTL format
  37. def triple(subject, predicate, object1):
  38. line = subject + ' ' + predicate + ' ' + object1
  39. return line
  40. def doublet(predicate, object1):
  41. line = ' ' + predicate + ' ' + object1
  42. return line
  43. def singlet(object1):
  44. line = ' ' + object1
  45. return line
  46. # Line endings in TTL format
  47. continueLine1 = ' ;\n'
  48. continueLine2 = ' ,\n'
  49. closeLine = ' .\n'
  50. def writeTTLHeader(output):
  51. output.write('@prefix ' + aspoCoords.prefix + ' ' + aspoCoords.uri + closeLine)
  52. output.write('@prefix ' + foafCoords.prefix + ' ' + foafCoords.uri + closeLine)
  53. output.write('@prefix ' + cidocCoords.prefix + ' ' + cidocCoords.uri + closeLine)
  54. output.write('@prefix ' + personCoords.prefix + ' ' + personCoords.uri + closeLine)
  55. output.write('@prefix ' + schemaCoords.prefix + ' ' + schemaCoords.uri + closeLine)
  56. output.write('@prefix ' + nsCoords.prefix + ' ' + nsCoords.uri + closeLine)
  57. output.write('@prefix ' + rdfsCoords.prefix + ' ' + rdfsCoords.uri + closeLine)
  58. output.write('@prefix ' + owlCoords.prefix + ' ' + owlCoords.uri + closeLine)
  59. output.write('\n')
  60. filePrefix = 'DATINI - onomastica '
  61. fileType = '- mani'
  62. max_entries = 10000000000000
  63. with open(import_dir + filePrefix + fileType + '.csv', newline="") as csv_file, open(
  64. export_dir + filePrefix + fileType + '.ttl', 'w') as output:
  65. reader = csv.DictReader(csv_file)
  66. writeTTLHeader(output)
  67. first = True
  68. ii = 0
  69. for row in reader:
  70. # The index ii is used to process a limited number of entries for testing purposes
  71. ii = ii + 1
  72. id_aspo = row['recordId']
  73. aspoPlaceHolder = aspoCoords.prefix + id_aspo
  74. line = triple(aspoPlaceHolder,
  75. nsCoords.prefix + 'type',
  76. cidocCoords.prefix + 'E74_Group') + closeLine
  77. output.write(line)
  78. line = triple(aspoPlaceHolder,
  79. nsCoords.prefix + 'type',
  80. foafCoords.prefix + 'Group') + closeLine
  81. output.write(line)
  82. line = triple(aspoPlaceHolder,
  83. cidocCoords.prefix + 'P2_has_type',
  84. '\"Gruppo scrittura lettera\"') + closeLine
  85. output.write(line)
  86. line = triple(aspoPlaceHolder,
  87. cidocCoords.prefix + 'P1_is_identified_by',
  88. aspoPlaceHolder + "_E42") + closeLine
  89. output.write(line)
  90. line = triple(aspoPlaceHolder + "_E42",
  91. nsCoords.prefix + 'type',
  92. cidocCoords.prefix + 'E42_Identifier') + closeLine
  93. output.write(line)
  94. line = triple(aspoPlaceHolder + "_E42",
  95. rdfsCoords.prefix + 'label',
  96. '\"' + id_aspo + '\"') + closeLine
  97. output.write(line)
  98. line = triple(aspoPlaceHolder,
  99. foafCoords.prefix + 'name',
  100. '\"' + row['nameEntry@normal'].replace("\"", "") + '\"') + closeLine
  101. output.write(line)
  102. line = triple(aspoPlaceHolder,
  103. rdfsCoords.prefix + 'label',
  104. '\"' + row['nameEntry@normal'].replace("\"", "") + '\"') + closeLine
  105. output.write(line)
  106. if row['ID prima persona'] != '':
  107. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID prima persona'].replace(' ', '_') + ">"
  108. line = triple(E13placeHolder,
  109. nsCoords.prefix + 'type',
  110. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  111. output.write(line)
  112. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  113. output.write(line)
  114. line = triple(E13placeHolder,
  115. rdfsCoords.prefix + 'label',
  116. '\"Relazione: ' + row['ID prima persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  117. output.write(line)
  118. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  119. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  120. output.write(line)
  121. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  122. output.write(line)
  123. id = row['ID prima persona']
  124. if re.match(r'IT-ASPO', id):
  125. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID prima persona']+ ">"
  126. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  127. output.write(line)
  128. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  129. output.write(line)
  130. else:
  131. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID prima persona'].replace(' ', '_').lower() + ">"
  132. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  133. output.write(line)
  134. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  135. output.write(line)
  136. line = triple(socionoid,
  137. rdfsCoords.prefix + 'label',
  138. '\"' + row['ID prima persona'] + '\"') + closeLine
  139. output.write(line)
  140. line = triple(socionoid,
  141. nsCoords.prefix + 'type',
  142. cidocCoords.prefix + 'E21_Person') + closeLine
  143. output.write(line)
  144. line = triple(socionoid,
  145. nsCoords.prefix + 'type',
  146. personCoords.prefix + 'Person') + closeLine
  147. output.write(line)
  148. line = triple(socionoid,
  149. nsCoords.prefix + 'type',
  150. foafCoords.prefix + 'person') + closeLine
  151. output.write(line)
  152. if row['ID seconda persona'] != '':
  153. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID seconda persona'].replace(' ', '_') + ">"
  154. line = triple(E13placeHolder,
  155. nsCoords.prefix + 'type',
  156. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  157. output.write(line)
  158. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  159. output.write(line)
  160. line = triple(E13placeHolder,
  161. rdfsCoords.prefix + 'label',
  162. '\"Relazione: ' + row['ID seconda persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  163. output.write(line)
  164. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  165. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  166. output.write(line)
  167. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  168. output.write(line)
  169. id = row['ID seconda persona']
  170. if re.match(r'IT-ASPO', id):
  171. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID seconda persona']+ ">"
  172. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  173. output.write(line)
  174. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  175. output.write(line)
  176. else:
  177. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID seconda persona'].replace(' ', '_').lower() + ">"
  178. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  179. output.write(line)
  180. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  181. output.write(line)
  182. line = triple(socionoid,
  183. rdfsCoords.prefix + 'label',
  184. '\"' + row['ID seconda persona'] + '\"') + closeLine
  185. output.write(line)
  186. line = triple(socionoid,
  187. nsCoords.prefix + 'type',
  188. cidocCoords.prefix + 'E21_Person') + closeLine
  189. output.write(line)
  190. line = triple(socionoid,
  191. nsCoords.prefix + 'type',
  192. personCoords.prefix + 'Person') + closeLine
  193. output.write(line)
  194. line = triple(socionoid,
  195. nsCoords.prefix + 'type',
  196. foafCoords.prefix + 'person') + closeLine
  197. output.write(line)
  198. if row['ID terza persona'] != '':
  199. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID terza persona'].replace(' ', '_') + ">"
  200. line = triple(E13placeHolder,
  201. nsCoords.prefix + 'type',
  202. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  203. output.write(line)
  204. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  205. output.write(line)
  206. line = triple(E13placeHolder,
  207. rdfsCoords.prefix + 'label',
  208. '\"Relazione: ' + row['ID terza persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  209. output.write(line)
  210. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  211. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  212. output.write(line)
  213. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  214. output.write(line)
  215. id = row['ID terza persona']
  216. if re.match(r'IT-ASPO', id):
  217. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID terza persona']+ ">"
  218. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  219. output.write(line)
  220. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  221. output.write(line)
  222. else:
  223. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID terza persona'].replace(' ', '_').lower() + ">"
  224. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  225. output.write(line)
  226. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  227. output.write(line)
  228. line = triple(socionoid,
  229. rdfsCoords.prefix + 'label',
  230. '\"' + row['ID terza persona'] + '\"') + closeLine
  231. output.write(line)
  232. line = triple(socionoid,
  233. nsCoords.prefix + 'type',
  234. cidocCoords.prefix + 'E21_Person') + closeLine
  235. output.write(line)
  236. line = triple(socionoid,
  237. nsCoords.prefix + 'type',
  238. personCoords.prefix + 'Person') + closeLine
  239. output.write(line)
  240. line = triple(socionoid,
  241. nsCoords.prefix + 'type',
  242. foafCoords.prefix + 'person') + closeLine
  243. output.write(line)
  244. if row['ID quarta persona'] != '':
  245. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID quarta persona'].replace(' ', '_') + ">"
  246. line = triple(E13placeHolder,
  247. nsCoords.prefix + 'type',
  248. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  249. output.write(line)
  250. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  251. output.write(line)
  252. line = triple(E13placeHolder,
  253. rdfsCoords.prefix + 'label',
  254. '\"Relazione: ' + row['ID quarta persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  255. output.write(line)
  256. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  257. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  258. output.write(line)
  259. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  260. output.write(line)
  261. id = row['ID quarta persona']
  262. if re.match(r'IT-ASPO', id):
  263. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID quarta persona']+ ">"
  264. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  265. output.write(line)
  266. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  267. output.write(line)
  268. else:
  269. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID quarta persona'].replace(' ', '_').lower() + ">"
  270. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  271. output.write(line)
  272. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  273. output.write(line)
  274. line = triple(socionoid,
  275. rdfsCoords.prefix + 'label',
  276. '\"' + row['ID quarta persona'] + '\"') + closeLine
  277. output.write(line)
  278. line = triple(socionoid,
  279. nsCoords.prefix + 'type',
  280. cidocCoords.prefix + 'E21_Person') + closeLine
  281. output.write(line)
  282. line = triple(socionoid,
  283. nsCoords.prefix + 'type',
  284. personCoords.prefix + 'Person') + closeLine
  285. output.write(line)
  286. line = triple(socionoid,
  287. nsCoords.prefix + 'type',
  288. foafCoords.prefix + 'person') + closeLine
  289. output.write(line)
  290. if row['ID quinta persona'] != '':
  291. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID quinta persona'].replace(' ', '_') + ">"
  292. line = triple(E13placeHolder,
  293. nsCoords.prefix + 'type',
  294. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  295. output.write(line)
  296. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  297. output.write(line)
  298. line = triple(E13placeHolder,
  299. rdfsCoords.prefix + 'label',
  300. '\"Relazione: ' + row['ID quinta persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  301. output.write(line)
  302. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  303. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  304. output.write(line)
  305. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  306. output.write(line)
  307. id = row['ID quinta persona']
  308. if re.match(r'IT-ASPO', id):
  309. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID quinta persona']+ ">"
  310. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  311. output.write(line)
  312. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  313. output.write(line)
  314. else:
  315. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID quinta persona'].replace(' ', '_').lower() + ">"
  316. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  317. output.write(line)
  318. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  319. output.write(line)
  320. line = triple(socionoid,
  321. rdfsCoords.prefix + 'label',
  322. '\"' + row['ID quinta persona'] + '\"') + closeLine
  323. output.write(line)
  324. line = triple(socionoid,
  325. nsCoords.prefix + 'type',
  326. cidocCoords.prefix + 'E21_Person') + closeLine
  327. output.write(line)
  328. line = triple(socionoid,
  329. nsCoords.prefix + 'type',
  330. personCoords.prefix + 'Person') + closeLine
  331. output.write(line)
  332. line = triple(socionoid,
  333. nsCoords.prefix + 'type',
  334. foafCoords.prefix + 'person') + closeLine
  335. output.write(line)
  336. if row['ID sesta persona'] != '':
  337. E13placeHolder = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['recordId'] + '_ML_' + row['ID sesta persona'].replace(' ', '_') + ">"
  338. line = triple(E13placeHolder,
  339. nsCoords.prefix + 'type',
  340. cidocCoords.prefix + 'E13_Attribute_Assignment') + closeLine
  341. output.write(line)
  342. line = triple(E13placeHolder, cidocCoords.prefix + 'P141_assigned', aspoPlaceHolder) + closeLine
  343. output.write(line)
  344. line = triple(E13placeHolder,
  345. rdfsCoords.prefix + 'label',
  346. '\"Relazione: ' + row['ID sesta persona'] + ' socio del gruppo scrittura lettera ' + row['recordId'] + '\"') + closeLine
  347. output.write(line)
  348. E55placeHolder = '<http://www.archiviodistato.prato.it/manilettere>'
  349. line = triple(E13placeHolder, cidocCoords.prefix + 'P42_assigned', E55placeHolder) + closeLine
  350. output.write(line)
  351. line = triple(E55placeHolder, rdfsCoords.prefix + 'label', '\"Mano lettera\"') + closeLine
  352. output.write(line)
  353. id = row['ID sesta persona']
  354. if re.match(r'IT-ASPO', id):
  355. socioid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID sesta persona']+ ">"
  356. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socioid) + closeLine
  357. output.write(line)
  358. line = triple(socioid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  359. output.write(line)
  360. else:
  361. socionoid = '<http://www.archiviodistato.prato.it/accedi-e-consulta/aspoMV001/scheda/' + row['ID sesta persona'].replace(' ', '_').lower() + ">"
  362. line = triple(aspoPlaceHolder, cidocCoords.prefix + 'P107_has_current_or_former_member', socionoid) + closeLine
  363. output.write(line)
  364. line = triple(socionoid, cidocCoords.prefix + 'P141_assigned', E13placeHolder) + closeLine
  365. output.write(line)
  366. line = triple(socionoid,
  367. rdfsCoords.prefix + 'label',
  368. '\"' + row['ID sesta persona'] + '\"') + closeLine
  369. output.write(line)
  370. line = triple(socionoid,
  371. nsCoords.prefix + 'type',
  372. cidocCoords.prefix + 'E21_Person') + closeLine
  373. output.write(line)
  374. line = triple(socionoid,
  375. nsCoords.prefix + 'type',
  376. personCoords.prefix + 'Person') + closeLine
  377. output.write(line)
  378. line = triple(socionoid,
  379. nsCoords.prefix + 'type',
  380. foafCoords.prefix + 'person') + closeLine
  381. output.write(line)
  382. output.write('\n')
  383. #
  384. # Limit number of entries processed (if desired)
  385. if (ii > max_entries):
  386. break