kreuzberg 3.16.0__py3-none-any.whl → 3.17.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (90) hide show
  1. kreuzberg/__init__.py +2 -0
  2. kreuzberg/_config.py +8 -9
  3. kreuzberg/_extractors/_base.py +0 -46
  4. kreuzberg/_extractors/_html.py +1 -1
  5. kreuzberg/_extractors/_pandoc.py +2 -2
  6. kreuzberg/_extractors/_pdf.py +4 -4
  7. kreuzberg/_gmft.py +2 -2
  8. kreuzberg/_mcp/server.py +1 -1
  9. kreuzberg/_mime_types.py +1 -1
  10. kreuzberg/_ocr/_easyocr.py +4 -9
  11. kreuzberg/_ocr/_paddleocr.py +1 -1
  12. kreuzberg/_ocr/_tesseract.py +15 -25
  13. kreuzberg/_token_reduction/__init__.py +11 -0
  14. kreuzberg/_token_reduction/_reducer.py +439 -0
  15. kreuzberg/_token_reduction/_stopwords.py +116 -0
  16. kreuzberg/_token_reduction/stopwords/af_stopwords.json +53 -0
  17. kreuzberg/_token_reduction/stopwords/ar_stopwords.json +482 -0
  18. kreuzberg/_token_reduction/stopwords/bg_stopwords.json +261 -0
  19. kreuzberg/_token_reduction/stopwords/bn_stopwords.json +400 -0
  20. kreuzberg/_token_reduction/stopwords/br_stopwords.json +1205 -0
  21. kreuzberg/_token_reduction/stopwords/ca_stopwords.json +280 -0
  22. kreuzberg/_token_reduction/stopwords/cs_stopwords.json +425 -0
  23. kreuzberg/_token_reduction/stopwords/da_stopwords.json +172 -0
  24. kreuzberg/_token_reduction/stopwords/de_stopwords.json +622 -0
  25. kreuzberg/_token_reduction/stopwords/el_stopwords.json +849 -0
  26. kreuzberg/_token_reduction/stopwords/en_stopwords.json +1300 -0
  27. kreuzberg/_token_reduction/stopwords/eo_stopwords.json +175 -0
  28. kreuzberg/_token_reduction/stopwords/es_stopwords.json +734 -0
  29. kreuzberg/_token_reduction/stopwords/et_stopwords.json +37 -0
  30. kreuzberg/_token_reduction/stopwords/eu_stopwords.json +100 -0
  31. kreuzberg/_token_reduction/stopwords/fa_stopwords.json +801 -0
  32. kreuzberg/_token_reduction/stopwords/fi_stopwords.json +849 -0
  33. kreuzberg/_token_reduction/stopwords/fr_stopwords.json +693 -0
  34. kreuzberg/_token_reduction/stopwords/ga_stopwords.json +111 -0
  35. kreuzberg/_token_reduction/stopwords/gl_stopwords.json +162 -0
  36. kreuzberg/_token_reduction/stopwords/gu_stopwords.json +226 -0
  37. kreuzberg/_token_reduction/stopwords/ha_stopwords.json +41 -0
  38. kreuzberg/_token_reduction/stopwords/he_stopwords.json +196 -0
  39. kreuzberg/_token_reduction/stopwords/hi_stopwords.json +227 -0
  40. kreuzberg/_token_reduction/stopwords/hr_stopwords.json +181 -0
  41. kreuzberg/_token_reduction/stopwords/hu_stopwords.json +791 -0
  42. kreuzberg/_token_reduction/stopwords/hy_stopwords.json +47 -0
  43. kreuzberg/_token_reduction/stopwords/id_stopwords.json +760 -0
  44. kreuzberg/_token_reduction/stopwords/it_stopwords.json +634 -0
  45. kreuzberg/_token_reduction/stopwords/ja_stopwords.json +136 -0
  46. kreuzberg/_token_reduction/stopwords/kn_stopwords.json +84 -0
  47. kreuzberg/_token_reduction/stopwords/ko_stopwords.json +681 -0
  48. kreuzberg/_token_reduction/stopwords/ku_stopwords.json +64 -0
  49. kreuzberg/_token_reduction/stopwords/la_stopwords.json +51 -0
  50. kreuzberg/_token_reduction/stopwords/lt_stopwords.json +476 -0
  51. kreuzberg/_token_reduction/stopwords/lv_stopwords.json +163 -0
  52. kreuzberg/_token_reduction/stopwords/ml_stopwords.json +11 -0
  53. kreuzberg/_token_reduction/stopwords/mr_stopwords.json +101 -0
  54. kreuzberg/_token_reduction/stopwords/ms_stopwords.json +477 -0
  55. kreuzberg/_token_reduction/stopwords/ne_stopwords.json +490 -0
  56. kreuzberg/_token_reduction/stopwords/nl_stopwords.json +415 -0
  57. kreuzberg/_token_reduction/stopwords/no_stopwords.json +223 -0
  58. kreuzberg/_token_reduction/stopwords/pl_stopwords.json +331 -0
  59. kreuzberg/_token_reduction/stopwords/pt_stopwords.json +562 -0
  60. kreuzberg/_token_reduction/stopwords/ro_stopwords.json +436 -0
  61. kreuzberg/_token_reduction/stopwords/ru_stopwords.json +561 -0
  62. kreuzberg/_token_reduction/stopwords/si_stopwords.json +193 -0
  63. kreuzberg/_token_reduction/stopwords/sk_stopwords.json +420 -0
  64. kreuzberg/_token_reduction/stopwords/sl_stopwords.json +448 -0
  65. kreuzberg/_token_reduction/stopwords/so_stopwords.json +32 -0
  66. kreuzberg/_token_reduction/stopwords/st_stopwords.json +33 -0
  67. kreuzberg/_token_reduction/stopwords/sv_stopwords.json +420 -0
  68. kreuzberg/_token_reduction/stopwords/sw_stopwords.json +76 -0
  69. kreuzberg/_token_reduction/stopwords/ta_stopwords.json +129 -0
  70. kreuzberg/_token_reduction/stopwords/te_stopwords.json +54 -0
  71. kreuzberg/_token_reduction/stopwords/th_stopwords.json +118 -0
  72. kreuzberg/_token_reduction/stopwords/tl_stopwords.json +149 -0
  73. kreuzberg/_token_reduction/stopwords/tr_stopwords.json +506 -0
  74. kreuzberg/_token_reduction/stopwords/uk_stopwords.json +75 -0
  75. kreuzberg/_token_reduction/stopwords/ur_stopwords.json +519 -0
  76. kreuzberg/_token_reduction/stopwords/vi_stopwords.json +647 -0
  77. kreuzberg/_token_reduction/stopwords/yo_stopwords.json +62 -0
  78. kreuzberg/_token_reduction/stopwords/zh_stopwords.json +796 -0
  79. kreuzberg/_token_reduction/stopwords/zu_stopwords.json +31 -0
  80. kreuzberg/_types.py +35 -3
  81. kreuzberg/_utils/_image_preprocessing.py +1 -1
  82. kreuzberg/_utils/_ref.py +14 -6
  83. kreuzberg/exceptions.py +0 -1
  84. kreuzberg/extraction.py +25 -9
  85. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/METADATA +4 -3
  86. kreuzberg-3.17.0.dist-info/RECORD +128 -0
  87. kreuzberg-3.16.0.dist-info/RECORD +0 -61
  88. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/WHEEL +0 -0
  89. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/entry_points.txt +0 -0
  90. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/licenses/LICENSE +0 -0
kreuzberg/_token_reduction/stopwords/ro_stopwords.json ADDED
@@ -0,0 +1,436 @@
1
+ [
2
+ "a",
3
+ "abia",
4
+ "acea",
5
+ "aceasta",
6
+ "această",
7
+ "aceea",
8
+ "aceeasi",
9
+ "acei",
10
+ "aceia",
11
+ "acel",
12
+ "acela",
13
+ "acelasi",
14
+ "acele",
15
+ "acelea",
16
+ "acest",
17
+ "acesta",
18
+ "aceste",
19
+ "acestea",
20
+ "acestei",
21
+ "acestia",
22
+ "acestui",
23
+ "aceşti",
24
+ "aceştia",
25
+ "acolo",
26
+ "acord",
27
+ "acum",
28
+ "adica",
29
+ "ai",
30
+ "aia",
31
+ "aibă",
32
+ "aici",
33
+ "aiurea",
34
+ "al",
35
+ "ala",
36
+ "alaturi",
37
+ "ale",
38
+ "alea",
39
+ "alt",
40
+ "alta",
41
+ "altceva",
42
+ "altcineva",
43
+ "alte",
44
+ "altfel",
45
+ "alti",
46
+ "altii",
47
+ "altul",
48
+ "am",
49
+ "anume",
50
+ "apoi",
51
+ "ar",
52
+ "are",
53
+ "as",
54
+ "asa",
55
+ "asemenea",
56
+ "asta",
57
+ "astazi",
58
+ "astea",
59
+ "astfel",
60
+ "astăzi",
61
+ "asupra",
62
+ "atare",
63
+ "atat",
64
+ "atata",
65
+ "atatea",
66
+ "atatia",
67
+ "ati",
68
+ "atit",
69
+ "atita",
70
+ "atitea",
71
+ "atitia",
72
+ "atunci",
73
+ "au",
74
+ "avea",
75
+ "avem",
76
+ "aveţi",
77
+ "avut",
78
+ "azi",
79
+ "aş",
80
+ "aşadar",
81
+ "aţi",
82
+ "b",
83
+ "ba",
84
+ "bine",
85
+ "bucur",
86
+ "bună",
87
+ "c",
88
+ "ca",
89
+ "cam",
90
+ "cand",
91
+ "capat",
92
+ "care",
93
+ "careia",
94
+ "carora",
95
+ "caruia",
96
+ "cat",
97
+ "catre",
98
+ "caut",
99
+ "ce",
100
+ "cea",
101
+ "ceea",
102
+ "cei",
103
+ "ceilalti",
104
+ "cel",
105
+ "cele",
106
+ "celor",
107
+ "ceva",
108
+ "chiar",
109
+ "ci",
110
+ "cinci",
111
+ "cind",
112
+ "cine",
113
+ "cineva",
114
+ "cit",
115
+ "cita",
116
+ "cite",
117
+ "citeva",
118
+ "citi",
119
+ "citiva",
120
+ "conform",
121
+ "contra",
122
+ "cu",
123
+ "cui",
124
+ "cum",
125
+ "cumva",
126
+ "curând",
127
+ "curînd",
128
+ "când",
129
+ "cât",
130
+ "câte",
131
+ "câtva",
132
+ "câţi",
133
+ "cînd",
134
+ "cît",
135
+ "cîte",
136
+ "cîtva",
137
+ "cîţi",
138
+ "că",
139
+ "căci",
140
+ "cărei",
141
+ "căror",
142
+ "cărui",
143
+ "către",
144
+ "d",
145
+ "da",
146
+ "daca",
147
+ "dacă",
148
+ "dar",
149
+ "dat",
150
+ "datorită",
151
+ "dată",
152
+ "dau",
153
+ "de",
154
+ "deasupra",
155
+ "deci",
156
+ "decit",
157
+ "degraba",
158
+ "deja",
159
+ "deoarece",
160
+ "departe",
161
+ "desi",
162
+ "despre",
163
+ "deşi",
164
+ "din",
165
+ "dinaintea",
166
+ "dintr",
167
+ "dintr-",
168
+ "dintre",
169
+ "doar",
170
+ "doi",
171
+ "doilea",
172
+ "două",
173
+ "drept",
174
+ "dupa",
175
+ "după",
176
+ "dă",
177
+ "e",
178
+ "ea",
179
+ "ei",
180
+ "el",
181
+ "ele",
182
+ "era",
183
+ "eram",
184
+ "este",
185
+ "eu",
186
+ "exact",
187
+ "eşti",
188
+ "f",
189
+ "face",
190
+ "fara",
191
+ "fata",
192
+ "fel",
193
+ "fi",
194
+ "fie",
195
+ "fiecare",
196
+ "fii",
197
+ "fim",
198
+ "fiu",
199
+ "fiţi",
200
+ "foarte",
201
+ "fost",
202
+ "frumos",
203
+ "fără",
204
+ "g",
205
+ "geaba",
206
+ "graţie",
207
+ "h",
208
+ "halbă",
209
+ "i",
210
+ "ia",
211
+ "iar",
212
+ "ieri",
213
+ "ii",
214
+ "il",
215
+ "imi",
216
+ "in",
217
+ "inainte",
218
+ "inapoi",
219
+ "inca",
220
+ "incit",
221
+ "insa",
222
+ "intr",
223
+ "intre",
224
+ "isi",
225
+ "iti",
226
+ "j",
227
+ "k",
228
+ "l",
229
+ "la",
230
+ "le",
231
+ "li",
232
+ "lor",
233
+ "lui",
234
+ "lângă",
235
+ "lîngă",
236
+ "m",
237
+ "ma",
238
+ "mai",
239
+ "mare",
240
+ "mea",
241
+ "mei",
242
+ "mele",
243
+ "mereu",
244
+ "meu",
245
+ "mi",
246
+ "mie",
247
+ "mine",
248
+ "mod",
249
+ "mult",
250
+ "multa",
251
+ "multe",
252
+ "multi",
253
+ "multă",
254
+ "mulţi",
255
+ "mulţumesc",
256
+ "mâine",
257
+ "mîine",
258
+ "mă",
259
+ "n",
260
+ "ne",
261
+ "nevoie",
262
+ "ni",
263
+ "nici",
264
+ "niciodata",
265
+ "nicăieri",
266
+ "nimeni",
267
+ "nimeri",
268
+ "nimic",
269
+ "niste",
270
+ "nişte",
271
+ "noastre",
272
+ "noastră",
273
+ "noi",
274
+ "noroc",
275
+ "nostri",
276
+ "nostru",
277
+ "nou",
278
+ "noua",
279
+ "nouă",
280
+ "noştri",
281
+ "nu",
282
+ "numai",
283
+ "o",
284
+ "opt",
285
+ "or",
286
+ "ori",
287
+ "oricare",
288
+ "orice",
289
+ "oricine",
290
+ "oricum",
291
+ "oricând",
292
+ "oricât",
293
+ "oricînd",
294
+ "oricît",
295
+ "oriunde",
296
+ "p",
297
+ "pai",
298
+ "parca",
299
+ "patra",
300
+ "patru",
301
+ "patrulea",
302
+ "pe",
303
+ "pentru",
304
+ "peste",
305
+ "pic",
306
+ "pina",
307
+ "plus",
308
+ "poate",
309
+ "pot",
310
+ "prea",
311
+ "prima",
312
+ "primul",
313
+ "prin",
314
+ "printr-",
315
+ "putini",
316
+ "puţin",
317
+ "puţina",
318
+ "puţină",
319
+ "până",
320
+ "pînă",
321
+ "r",
322
+ "rog",
323
+ "s",
324
+ "sa",
325
+ "sa-mi",
326
+ "sa-ti",
327
+ "sai",
328
+ "sale",
329
+ "sau",
330
+ "se",
331
+ "si",
332
+ "sint",
333
+ "sintem",
334
+ "spate",
335
+ "spre",
336
+ "sub",
337
+ "sunt",
338
+ "suntem",
339
+ "sunteţi",
340
+ "sus",
341
+ "sută",
342
+ "sînt",
343
+ "sîntem",
344
+ "sînteţi",
345
+ "să",
346
+ "săi",
347
+ "său",
348
+ "t",
349
+ "ta",
350
+ "tale",
351
+ "te",
352
+ "ti",
353
+ "timp",
354
+ "tine",
355
+ "toata",
356
+ "toate",
357
+ "toată",
358
+ "tocmai",
359
+ "tot",
360
+ "toti",
361
+ "totul",
362
+ "totusi",
363
+ "totuşi",
364
+ "toţi",
365
+ "trei",
366
+ "treia",
367
+ "treilea",
368
+ "tu",
369
+ "tuturor",
370
+ "tăi",
371
+ "tău",
372
+ "u",
373
+ "ul",
374
+ "ului",
375
+ "un",
376
+ "una",
377
+ "unde",
378
+ "undeva",
379
+ "unei",
380
+ "uneia",
381
+ "unele",
382
+ "uneori",
383
+ "unii",
384
+ "unor",
385
+ "unora",
386
+ "unu",
387
+ "unui",
388
+ "unuia",
389
+ "unul",
390
+ "v",
391
+ "va",
392
+ "vi",
393
+ "voastre",
394
+ "voastră",
395
+ "voi",
396
+ "vom",
397
+ "vor",
398
+ "vostru",
399
+ "vouă",
400
+ "voştri",
401
+ "vreme",
402
+ "vreo",
403
+ "vreun",
404
+ "vă",
405
+ "x",
406
+ "z",
407
+ "zece",
408
+ "zero",
409
+ "zi",
410
+ "zice",
411
+ "îi",
412
+ "îl",
413
+ "îmi",
414
+ "împotriva",
415
+ "în",
416
+ "înainte",
417
+ "înaintea",
418
+ "încotro",
419
+ "încât",
420
+ "încît",
421
+ "între",
422
+ "întrucât",
423
+ "întrucît",
424
+ "îţi",
425
+ "ăla",
426
+ "ălea",
427
+ "ăsta",
428
+ "ăstea",
429
+ "ăştia",
430
+ "şapte",
431
+ "şase",
432
+ "şi",
433
+ "ştiu",
434
+ "ţi",
435
+ "ţie"
436
+ ]