kreuzberg 3.16.0__py3-none-any.whl → 3.17.1__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 (91) 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/_language_detection.py +16 -39
  9. kreuzberg/_mcp/server.py +1 -1
  10. kreuzberg/_mime_types.py +1 -1
  11. kreuzberg/_ocr/_easyocr.py +4 -9
  12. kreuzberg/_ocr/_paddleocr.py +1 -1
  13. kreuzberg/_ocr/_tesseract.py +15 -25
  14. kreuzberg/_token_reduction/__init__.py +11 -0
  15. kreuzberg/_token_reduction/_reducer.py +439 -0
  16. kreuzberg/_token_reduction/_stopwords.py +116 -0
  17. kreuzberg/_token_reduction/stopwords/af_stopwords.json +53 -0
  18. kreuzberg/_token_reduction/stopwords/ar_stopwords.json +482 -0
  19. kreuzberg/_token_reduction/stopwords/bg_stopwords.json +261 -0
  20. kreuzberg/_token_reduction/stopwords/bn_stopwords.json +400 -0
  21. kreuzberg/_token_reduction/stopwords/br_stopwords.json +1205 -0
  22. kreuzberg/_token_reduction/stopwords/ca_stopwords.json +280 -0
  23. kreuzberg/_token_reduction/stopwords/cs_stopwords.json +425 -0
  24. kreuzberg/_token_reduction/stopwords/da_stopwords.json +172 -0
  25. kreuzberg/_token_reduction/stopwords/de_stopwords.json +622 -0
  26. kreuzberg/_token_reduction/stopwords/el_stopwords.json +849 -0
  27. kreuzberg/_token_reduction/stopwords/en_stopwords.json +1300 -0
  28. kreuzberg/_token_reduction/stopwords/eo_stopwords.json +175 -0
  29. kreuzberg/_token_reduction/stopwords/es_stopwords.json +734 -0
  30. kreuzberg/_token_reduction/stopwords/et_stopwords.json +37 -0
  31. kreuzberg/_token_reduction/stopwords/eu_stopwords.json +100 -0
  32. kreuzberg/_token_reduction/stopwords/fa_stopwords.json +801 -0
  33. kreuzberg/_token_reduction/stopwords/fi_stopwords.json +849 -0
  34. kreuzberg/_token_reduction/stopwords/fr_stopwords.json +693 -0
  35. kreuzberg/_token_reduction/stopwords/ga_stopwords.json +111 -0
  36. kreuzberg/_token_reduction/stopwords/gl_stopwords.json +162 -0
  37. kreuzberg/_token_reduction/stopwords/gu_stopwords.json +226 -0
  38. kreuzberg/_token_reduction/stopwords/ha_stopwords.json +41 -0
  39. kreuzberg/_token_reduction/stopwords/he_stopwords.json +196 -0
  40. kreuzberg/_token_reduction/stopwords/hi_stopwords.json +227 -0
  41. kreuzberg/_token_reduction/stopwords/hr_stopwords.json +181 -0
  42. kreuzberg/_token_reduction/stopwords/hu_stopwords.json +791 -0
  43. kreuzberg/_token_reduction/stopwords/hy_stopwords.json +47 -0
  44. kreuzberg/_token_reduction/stopwords/id_stopwords.json +760 -0
  45. kreuzberg/_token_reduction/stopwords/it_stopwords.json +634 -0
  46. kreuzberg/_token_reduction/stopwords/ja_stopwords.json +136 -0
  47. kreuzberg/_token_reduction/stopwords/kn_stopwords.json +84 -0
  48. kreuzberg/_token_reduction/stopwords/ko_stopwords.json +681 -0
  49. kreuzberg/_token_reduction/stopwords/ku_stopwords.json +64 -0
  50. kreuzberg/_token_reduction/stopwords/la_stopwords.json +51 -0
  51. kreuzberg/_token_reduction/stopwords/lt_stopwords.json +476 -0
  52. kreuzberg/_token_reduction/stopwords/lv_stopwords.json +163 -0
  53. kreuzberg/_token_reduction/stopwords/ml_stopwords.json +11 -0
  54. kreuzberg/_token_reduction/stopwords/mr_stopwords.json +101 -0
  55. kreuzberg/_token_reduction/stopwords/ms_stopwords.json +477 -0
  56. kreuzberg/_token_reduction/stopwords/ne_stopwords.json +490 -0
  57. kreuzberg/_token_reduction/stopwords/nl_stopwords.json +415 -0
  58. kreuzberg/_token_reduction/stopwords/no_stopwords.json +223 -0
  59. kreuzberg/_token_reduction/stopwords/pl_stopwords.json +331 -0
  60. kreuzberg/_token_reduction/stopwords/pt_stopwords.json +562 -0
  61. kreuzberg/_token_reduction/stopwords/ro_stopwords.json +436 -0
  62. kreuzberg/_token_reduction/stopwords/ru_stopwords.json +561 -0
  63. kreuzberg/_token_reduction/stopwords/si_stopwords.json +193 -0
  64. kreuzberg/_token_reduction/stopwords/sk_stopwords.json +420 -0
  65. kreuzberg/_token_reduction/stopwords/sl_stopwords.json +448 -0
  66. kreuzberg/_token_reduction/stopwords/so_stopwords.json +32 -0
  67. kreuzberg/_token_reduction/stopwords/st_stopwords.json +33 -0
  68. kreuzberg/_token_reduction/stopwords/sv_stopwords.json +420 -0
  69. kreuzberg/_token_reduction/stopwords/sw_stopwords.json +76 -0
  70. kreuzberg/_token_reduction/stopwords/ta_stopwords.json +129 -0
  71. kreuzberg/_token_reduction/stopwords/te_stopwords.json +54 -0
  72. kreuzberg/_token_reduction/stopwords/th_stopwords.json +118 -0
  73. kreuzberg/_token_reduction/stopwords/tl_stopwords.json +149 -0
  74. kreuzberg/_token_reduction/stopwords/tr_stopwords.json +506 -0
  75. kreuzberg/_token_reduction/stopwords/uk_stopwords.json +75 -0
  76. kreuzberg/_token_reduction/stopwords/ur_stopwords.json +519 -0
  77. kreuzberg/_token_reduction/stopwords/vi_stopwords.json +647 -0
  78. kreuzberg/_token_reduction/stopwords/yo_stopwords.json +62 -0
  79. kreuzberg/_token_reduction/stopwords/zh_stopwords.json +796 -0
  80. kreuzberg/_token_reduction/stopwords/zu_stopwords.json +31 -0
  81. kreuzberg/_types.py +50 -9
  82. kreuzberg/_utils/_image_preprocessing.py +1 -1
  83. kreuzberg/_utils/_ref.py +14 -6
  84. kreuzberg/exceptions.py +0 -1
  85. kreuzberg/extraction.py +33 -10
  86. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.1.dist-info}/METADATA +6 -5
  87. kreuzberg-3.17.1.dist-info/RECORD +128 -0
  88. kreuzberg-3.16.0.dist-info/RECORD +0 -61
  89. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.1.dist-info}/WHEEL +0 -0
  90. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.1.dist-info}/entry_points.txt +0 -0
  91. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.1.dist-info}/licenses/LICENSE +0 -0
kreuzberg/_token_reduction/stopwords/pl_stopwords.json ADDED
@@ -0,0 +1,331 @@
1
+ [
2
+ "a",
3
+ "aby",
4
+ "ach",
5
+ "acz",
6
+ "aczkolwiek",
7
+ "aj",
8
+ "albo",
9
+ "ale",
10
+ "ależ",
11
+ "ani",
12
+ "aż",
13
+ "bardziej",
14
+ "bardzo",
15
+ "bez",
16
+ "bo",
17
+ "bowiem",
18
+ "by",
19
+ "byli",
20
+ "bym",
21
+ "bynajmniej",
22
+ "być",
23
+ "był",
24
+ "była",
25
+ "było",
26
+ "były",
27
+ "będzie",
28
+ "będą",
29
+ "cali",
30
+ "cała",
31
+ "cały",
32
+ "chce",
33
+ "choć",
34
+ "ci",
35
+ "ciebie",
36
+ "cię",
37
+ "co",
38
+ "cokolwiek",
39
+ "coraz",
40
+ "coś",
41
+ "czasami",
42
+ "czasem",
43
+ "czemu",
44
+ "czy",
45
+ "czyli",
46
+ "często",
47
+ "daleko",
48
+ "dla",
49
+ "dlaczego",
50
+ "dlatego",
51
+ "do",
52
+ "dobrze",
53
+ "dokąd",
54
+ "dość",
55
+ "dr",
56
+ "dużo",
57
+ "dwa",
58
+ "dwaj",
59
+ "dwie",
60
+ "dwoje",
61
+ "dzisiaj",
62
+ "dziś",
63
+ "gdy",
64
+ "gdyby",
65
+ "gdyż",
66
+ "gdzie",
67
+ "gdziekolwiek",
68
+ "gdzieś",
69
+ "go",
70
+ "godz",
71
+ "hab",
72
+ "i",
73
+ "ich",
74
+ "ii",
75
+ "iii",
76
+ "ile",
77
+ "im",
78
+ "inna",
79
+ "inne",
80
+ "inny",
81
+ "innych",
82
+ "inż",
83
+ "iv",
84
+ "ix",
85
+ "iż",
86
+ "ja",
87
+ "jak",
88
+ "jakaś",
89
+ "jakby",
90
+ "jaki",
91
+ "jakichś",
92
+ "jakie",
93
+ "jakiś",
94
+ "jakiż",
95
+ "jakkolwiek",
96
+ "jako",
97
+ "jakoś",
98
+ "je",
99
+ "jeden",
100
+ "jedna",
101
+ "jednak",
102
+ "jednakże",
103
+ "jedno",
104
+ "jednym",
105
+ "jedynie",
106
+ "jego",
107
+ "jej",
108
+ "jemu",
109
+ "jest",
110
+ "jestem",
111
+ "jeszcze",
112
+ "jeśli",
113
+ "jeżeli",
114
+ "już",
115
+ "ją",
116
+ "każdy",
117
+ "kiedy",
118
+ "kierunku",
119
+ "kilka",
120
+ "kilku",
121
+ "kimś",
122
+ "kto",
123
+ "ktokolwiek",
124
+ "ktoś",
125
+ "która",
126
+ "które",
127
+ "którego",
128
+ "której",
129
+ "który",
130
+ "których",
131
+ "którym",
132
+ "którzy",
133
+ "ku",
134
+ "lat",
135
+ "lecz",
136
+ "lub",
137
+ "ma",
138
+ "mają",
139
+ "mam",
140
+ "mamy",
141
+ "mało",
142
+ "mgr",
143
+ "mi",
144
+ "miał",
145
+ "mimo",
146
+ "między",
147
+ "mnie",
148
+ "mną",
149
+ "mogą",
150
+ "moi",
151
+ "moim",
152
+ "moja",
153
+ "moje",
154
+ "może",
155
+ "możliwe",
156
+ "można",
157
+ "mu",
158
+ "musi",
159
+ "my",
160
+ "mój",
161
+ "na",
162
+ "nad",
163
+ "nam",
164
+ "nami",
165
+ "nas",
166
+ "nasi",
167
+ "nasz",
168
+ "nasza",
169
+ "nasze",
170
+ "naszego",
171
+ "naszych",
172
+ "natomiast",
173
+ "natychmiast",
174
+ "nawet",
175
+ "nic",
176
+ "nich",
177
+ "nie",
178
+ "niech",
179
+ "niego",
180
+ "niej",
181
+ "niemu",
182
+ "nigdy",
183
+ "nim",
184
+ "nimi",
185
+ "nią",
186
+ "niż",
187
+ "no",
188
+ "nowe",
189
+ "np",
190
+ "nr",
191
+ "o",
192
+ "o.o.",
193
+ "obok",
194
+ "od",
195
+ "ok",
196
+ "około",
197
+ "on",
198
+ "ona",
199
+ "one",
200
+ "oni",
201
+ "ono",
202
+ "oraz",
203
+ "oto",
204
+ "owszem",
205
+ "pan",
206
+ "pana",
207
+ "pani",
208
+ "pl",
209
+ "po",
210
+ "pod",
211
+ "podczas",
212
+ "pomimo",
213
+ "ponad",
214
+ "ponieważ",
215
+ "powinien",
216
+ "powinna",
217
+ "powinni",
218
+ "powinno",
219
+ "poza",
220
+ "prawie",
221
+ "prof",
222
+ "przecież",
223
+ "przed",
224
+ "przede",
225
+ "przedtem",
226
+ "przez",
227
+ "przy",
228
+ "raz",
229
+ "razie",
230
+ "roku",
231
+ "również",
232
+ "sam",
233
+ "sama",
234
+ "się",
235
+ "skąd",
236
+ "sobie",
237
+ "sobą",
238
+ "sposób",
239
+ "swoje",
240
+ "są",
241
+ "ta",
242
+ "tak",
243
+ "taka",
244
+ "taki",
245
+ "takich",
246
+ "takie",
247
+ "także",
248
+ "tam",
249
+ "te",
250
+ "tego",
251
+ "tej",
252
+ "tel",
253
+ "temu",
254
+ "ten",
255
+ "teraz",
256
+ "też",
257
+ "to",
258
+ "tobie",
259
+ "tobą",
260
+ "toteż",
261
+ "totobą",
262
+ "trzeba",
263
+ "tu",
264
+ "tutaj",
265
+ "twoi",
266
+ "twoim",
267
+ "twoja",
268
+ "twoje",
269
+ "twym",
270
+ "twój",
271
+ "ty",
272
+ "tych",
273
+ "tylko",
274
+ "tym",
275
+ "tys",
276
+ "tzw",
277
+ "tę",
278
+ "u",
279
+ "ul",
280
+ "vi",
281
+ "vii",
282
+ "viii",
283
+ "vol",
284
+ "w",
285
+ "wam",
286
+ "wami",
287
+ "was",
288
+ "wasi",
289
+ "wasz",
290
+ "wasza",
291
+ "wasze",
292
+ "we",
293
+ "według",
294
+ "wie",
295
+ "wiele",
296
+ "wielu",
297
+ "więc",
298
+ "więcej",
299
+ "wszyscy",
300
+ "wszystkich",
301
+ "wszystkie",
302
+ "wszystkim",
303
+ "wszystko",
304
+ "wtedy",
305
+ "www",
306
+ "wy",
307
+ "właśnie",
308
+ "wśród",
309
+ "xi",
310
+ "xii",
311
+ "xiii",
312
+ "xiv",
313
+ "xv",
314
+ "z",
315
+ "za",
316
+ "zapewne",
317
+ "zawsze",
318
+ "zaś",
319
+ "ze",
320
+ "zeznowu",
321
+ "znowu",
322
+ "znów",
323
+ "został",
324
+ "zł",
325
+ "żaden",
326
+ "żadna",
327
+ "żadne",
328
+ "żadnych",
329
+ "że",
330
+ "żeby"
331
+ ]