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/ja_stopwords.json ADDED
@@ -0,0 +1,136 @@
1
+ [
2
+ "あそこ",
3
+ "あっ",
4
+ "あの",
5
+ "あのかた",
6
+ "あの人",
7
+ "あり",
8
+ "あります",
9
+ "ある",
10
+ "あれ",
11
+ "い",
12
+ "いう",
13
+ "います",
14
+ "いる",
15
+ "う",
16
+ "うち",
17
+ "え",
18
+ "お",
19
+ "および",
20
+ "おり",
21
+ "おります",
22
+ "か",
23
+ "かつて",
24
+ "から",
25
+ "が",
26
+ "き",
27
+ "ここ",
28
+ "こちら",
29
+ "こと",
30
+ "この",
31
+ "これ",
32
+ "これら",
33
+ "さ",
34
+ "さらに",
35
+ "し",
36
+ "しかし",
37
+ "する",
38
+ "ず",
39
+ "せ",
40
+ "せる",
41
+ "そこ",
42
+ "そして",
43
+ "その",
44
+ "その他",
45
+ "その後",
46
+ "それ",
47
+ "それぞれ",
48
+ "それで",
49
+ "た",
50
+ "ただし",
51
+ "たち",
52
+ "ため",
53
+ "たり",
54
+ "だ",
55
+ "だっ",
56
+ "だれ",
57
+ "つ",
58
+ "て",
59
+ "で",
60
+ "でき",
61
+ "できる",
62
+ "です",
63
+ "では",
64
+ "でも",
65
+ "と",
66
+ "という",
67
+ "といった",
68
+ "とき",
69
+ "ところ",
70
+ "として",
71
+ "とともに",
72
+ "とも",
73
+ "と共に",
74
+ "どこ",
75
+ "どの",
76
+ "な",
77
+ "ない",
78
+ "なお",
79
+ "なかっ",
80
+ "ながら",
81
+ "なく",
82
+ "なっ",
83
+ "など",
84
+ "なに",
85
+ "なら",
86
+ "なり",
87
+ "なる",
88
+ "なん",
89
+ "に",
90
+ "において",
91
+ "における",
92
+ "について",
93
+ "にて",
94
+ "によって",
95
+ "により",
96
+ "による",
97
+ "に対して",
98
+ "に対する",
99
+ "に関する",
100
+ "の",
101
+ "ので",
102
+ "のみ",
103
+ "は",
104
+ "ば",
105
+ "へ",
106
+ "ほか",
107
+ "ほとんど",
108
+ "ほど",
109
+ "ます",
110
+ "また",
111
+ "または",
112
+ "まで",
113
+ "も",
114
+ "もの",
115
+ "ものの",
116
+ "や",
117
+ "よう",
118
+ "より",
119
+ "ら",
120
+ "られ",
121
+ "られる",
122
+ "れ",
123
+ "れる",
124
+ "を",
125
+ "ん",
126
+ "何",
127
+ "及び",
128
+ "彼",
129
+ "彼女",
130
+ "我々",
131
+ "特に",
132
+ "私",
133
+ "私達",
134
+ "貴方",
135
+ "貴方方"
136
+ ]
kreuzberg/_token_reduction/stopwords/kn_stopwords.json ADDED
@@ -0,0 +1,84 @@
1
+ [
2
+ "ಅಂತ",
3
+ "ಅಥವಾ",
4
+ "ಅದಕ್ಕೆ",
5
+ "ಅದನ್ನು",
6
+ "ಅದು",
7
+ "ಅದೇ",
8
+ "ಅಲ್ಲ",
9
+ "ಅಲ್ಲದೆ",
10
+ "ಅವರ",
11
+ "ಅವರನ್ನು",
12
+ "ಅವರಿಗೆ",
13
+ "ಅವರು",
14
+ "ಅಷ್ಟೇ",
15
+ "ಆ",
16
+ "ಆಗ",
17
+ "ಆಗಿ",
18
+ "ಆದರೆ",
19
+ "ಇತ್ತು",
20
+ "ಇದಕ್ಕೆ",
21
+ "ಇದನ್ನು",
22
+ "ಇದರ",
23
+ "ಇದರಿಂದ",
24
+ "ಇದು",
25
+ "ಇದೆ",
26
+ "ಇದೇ",
27
+ "ಇನ್ನು",
28
+ "ಇನ್ನೂ",
29
+ "ಇರುವ",
30
+ "ಇಲ್ಲ",
31
+ "ಇಲ್ಲಿ",
32
+ "ಈ",
33
+ "ಈಗ",
34
+ "ಈಗಾಗಲೇ",
35
+ "ಎಂದರು",
36
+ "ಎಂದರೆ",
37
+ "ಎಂದು",
38
+ "ಎಂಬ",
39
+ "ಎಂಬುದು",
40
+ "ಎಲ್ಲ",
41
+ "ಎಲ್ಲಾ",
42
+ "ಎಷ್ಟು",
43
+ "ಏನು",
44
+ "ಕಳೆದ",
45
+ "ಕೂಡ",
46
+ "ಕೆಲವು",
47
+ "ಕೇವಲ",
48
+ "ಜೊತೆಗೆ",
49
+ "ತನ್ನ",
50
+ "ತಮ್ಮ",
51
+ "ತುಂಬಾ",
52
+ "ನಡೆದ",
53
+ "ನಡೆದಿದೆ",
54
+ "ನನಗೆ",
55
+ "ನನ್ನ",
56
+ "ನನ್ನನ್ನು",
57
+ "ನಮ್ಮ",
58
+ "ನಾನು",
59
+ "ನಾವು",
60
+ "ನಿಮ್ಮ",
61
+ "ನೀಡಿದರು",
62
+ "ನೀಡಿದೆ",
63
+ "ನೀಡಿದ್ದಾರೆ",
64
+ "ನೀವು",
65
+ "ಬಗ್ಗೆ",
66
+ "ಬಳಿಕ",
67
+ "ಬೇರೆ",
68
+ "ಮತ್ತು",
69
+ "ಮತ್ತೆ",
70
+ "ಮಾಡಿ",
71
+ "ಮಾತ್ರ",
72
+ "ಮೂಲಕ",
73
+ "ಮೇಲೆ",
74
+ "ಯಾವ",
75
+ "ಯಾವುದೇ",
76
+ "ಹಲವು",
77
+ "ಹಾಗಾಗಿ",
78
+ "ಹಾಗೂ",
79
+ "ಹೀಗಾಗಿ",
80
+ "ಹೀಗೆ",
81
+ "ಹೆಚ್ಚು",
82
+ "ಹೇಳಿ",
83
+ "ಹೇಳಿದ್ದಾರೆ"
84
+ ]