Hugging Face's logo

Datasets:
aisingapore
/
WangchanLION-Curated

Tasks:
Fill-Mask
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
Thai
Size:
100K - 1M
ArXiv:
Tags:
legal
finance
medical
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files Community
1
mrpeerat commited on
Commit
9d0710c
·
verified ·
1 Parent(s): 513b59d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +141 -55
README.md CHANGED
@@ -1,55 +1,141 @@
1
- ---
2
- license: cc
3
- dataset_info:
4
- features:
5
- - name: id
6
- dtype: string
7
- - name: text
8
- dtype: string
9
- - name: source
10
- dtype: string
11
- - name: metadata
12
- struct:
13
- - name: Filename
14
- dtype: string
15
- - name: author
16
- dtype: string
17
- - name: channel_name
18
- dtype: string
19
- - name: domain
20
- dtype: string
21
- - name: is_subtitle_generated
22
- dtype: string
23
- - name: license
24
- dtype: string
25
- - name: provenance
26
- dtype: string
27
- - name: revid
28
- dtype: string
29
- - name: src
30
- dtype: string
31
- - name: ticker
32
- dtype: string
33
- - name: title
34
- dtype: string
35
- - name: url
36
- dtype: string
37
- - name: year
38
- dtype: float64
39
- splits:
40
- - name: train
41
- num_bytes: 8513541142
42
- num_examples: 397488
43
- - name: validation
44
- num_bytes: 83508453
45
- num_examples: 4044
46
- download_size: 2639043294
47
- dataset_size: 8597049595
48
- configs:
49
- - config_name: default
50
- data_files:
51
- - split: train
52
- path: data/train-*
53
- - split: validation
54
- path: data/validation-*
55
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc
3
+ dataset_info:
4
+ features:
5
+ - name: id
6
+ dtype: string
7
+ - name: text
8
+ dtype: string
9
+ - name: source
10
+ dtype: string
11
+ - name: metadata
12
+ struct:
13
+ - name: Filename
14
+ dtype: string
15
+ - name: author
16
+ dtype: string
17
+ - name: channel_name
18
+ dtype: string
19
+ - name: domain
20
+ dtype: string
21
+ - name: is_subtitle_generated
22
+ dtype: string
23
+ - name: license
24
+ dtype: string
25
+ - name: provenance
26
+ dtype: string
27
+ - name: revid
28
+ dtype: string
29
+ - name: src
30
+ dtype: string
31
+ - name: ticker
32
+ dtype: string
33
+ - name: title
34
+ dtype: string
35
+ - name: url
36
+ dtype: string
37
+ - name: year
38
+ dtype: float64
39
+ splits:
40
+ - name: train
41
+ num_bytes: 8513541142
42
+ num_examples: 397488
43
+ - name: validation
44
+ num_bytes: 83508453
45
+ num_examples: 4044
46
+ download_size: 2639043294
47
+ dataset_size: 8597049595
48
+ configs:
49
+ - config_name: default
50
+ data_files:
51
+ - split: train
52
+ path: data/train-*
53
+ - split: validation
54
+ path: data/validation-*
55
+ ---
56
+
57
+ We have collected additional Thai text that is unlikely to be included in the common crawl from various sources. The total number of documents collected is as follows:425,304 documents, we deduplication these noncc documents to later divide them into train sets for further web data and validation set.
58
+
59
+ The following table shows the analysis of the data after deduplication.
60
+
61
+ | type | Source | Number of documents | Number of words | Cultural? | Human-craft? | Evaluation topics? |
62
+ |----------------------|------------------------------------------------------------|---------------------|-----------------|:---------:|:------------:|:-----------------------------:|
63
+ | encyclopedic | th.wikibooks.org | 1,179 | 756,175 | No | | Common safety |
64
+ | | th.wikipedia.org | 162,189 | 74,151,797 | No | | Common safety |
65
+ | | th.wikiquote.org | 929 | 156,910 | No | | Common safety |
66
+ | | th.wikisource.org | 1,890 | 5,601,908 | No | | Common safety |
67
+ | finance | airesearch/cmdf_vistec | 86,813 | 348,189,390 | No | | Common safety |
68
+ | government document | lift catalog data | 1 | 6,931 | No | Yes/No | Common safety |
69
+ | | data.go.th | 9,488 | 2,766,950 | No | No | Common safety |
70
+ | | envilink.go.th | 1 | 10,659 | | | |
71
+ | | government data catalog smart plus | 427 | 5,290,950 | Yes | Yes/No | Common safety/ Cross-lingual |
72
+ | | https://ratchakitcha.soc.go.th | 59,744 | 175,031,218 | Yes | Yes | Culture safety/ Cross-lingual |
73
+ | | nakhoratchasima data catalog | 1 | 68,000 | Yes | Yes | Country safety/ Cross-lingual |
74
+ | | opdc data portal | 3,148 | 598,829 | | | |
75
+ | | open-d | 7 | 284,100 | | | |
76
+ | | royal thai government | 1 | 41,356 | | | |
77
+ | | National Economic and Social Development Board | 1 | 37,663 | | | |
78
+ | | pythainlp/thailand-policy-statements | 60 | 226,087 | | | |
79
+ | legal | pythainlp/thai-cc-license | 6 | 50,727 | | | |
80
+ | | pythainlp/thai-constitution-corpus | 20 | 444,313 | | | |
81
+ | | pythainlp/thailaw-v1.0 | 52,317 | 79,715,118 | | | |
82
+ | academic literature | government data catalog smart plus | 427 | 5,290,950 | | | |
83
+ | | openbase.in.th | 4,173 | 165,909,425 | | | |
84
+ | | platform for social empowerment and transformation | 90 | 209,716 | | | |
85
+ | | pythainlp/thai-it-books | 7 | 174,644 | | | |
86
+ | | pythainlp/thai-tnhc2-books | 353 | 22,002,703 | | | |
87
+ | | pythainlp/tlcv2.0_oa | 361 | 2,970,463 | | | |
88
+ | | TDRI | 25 | 2,801,106 | | | |
89
+ | | Bangkok Open Data | 10 | 2,334 | | | |
90
+ | | Open educational resources repository | 14 | 47,951 | | | |
91
+ | | CMU Journal of Law and Social Sciences | 47 | 37,976 | | | |
92
+ | | E-journal of education studies, Burapha University | 68 | 59,137 | | | |
93
+ | | Chulalongkorn University Law Journal | 64 | 46,425 | | | |
94
+ | | Lanna Journal of Health Promotion and Environmental Health | 53 | 52,320 | | | |
95
+ | | Journal of Educational Studies, Burapha University | 64 | 56,320 | | | |
96
+ | | Journal of Yanasangwon Research Institute | 65 | 47,286 | | | |
97
+ | | Journal of Food and Drug Administration | 79 | 115,541 | | | |
98
+ | | https://github.com/kongruksiamza/ebook-for-education | 8 | 83,432 | | | |
99
+ | | social technology institute | 3 | 11,152 | | | |
100
+ | youtube | youtube | 17,826 | 46,613,632 | | | |
101
+
102
+ Documents from some sources cannot be directly used or easily processed. It is necessary to use text extraction technology (OCR) to extract the text due to the document format in PDF. The table below shows the number and proportion of documents that required OCR.
103
+
104
+ | Source | Number of documents | Number of documents required for PCR | Percentage of documents requiring OCR |
105
+ |--------------------------------------:|--------------------:|-------------------------------------:|--------------------------------------:|
106
+ | data.go.th | 9488 | 18 | 0.189713 |
107
+ | government data catalog smart plus | 427 | 198 | 46.370023 |
108
+ | ebook construction | 8 | 8 | 100 |
109
+ | openbase.in.th | 4173 | 3443 | 82.50659 |
110
+ | opendata.nesdc.go.th | 7 | 3 | 42.857143 |
111
+ | royal thai government | 1 | 1 | 100 |
112
+ | Open educational resources repository | 14 | 2 | 14.285714 |
113
+
114
+ The model used for text extraction ishttps://github.com/VikParuchuri/marker
115
+ In the future, you should try VLM, such as:https://olmocr.allenai.org/Or Typhoon2 Vision
116
+
117
+ Data from various sources can be classified into 6 types as shown in the table below.
118
+
119
+ | Domain | count | proportion |
120
+ |-------------:|-------:|-----------:|
121
+ | Encyclopedic | 166187 | 41.34 |
122
+ | Finance | 86813 | 21.59 |
123
+ | Government | 72,879 | 18.13 |
124
+ | Legal | 52,343 | 13.02 |
125
+ | YouTube | 17,837 | 4.43 |
126
+ | Education | 5,911 | 1.47 |
127
+
128
+ The additional documents we collect are confirmed to be open source and have a license to allow for redistribution, with the copyright share as shown in the table below.
129
+
130
+ | license | count | proportion |
131
+ |----------------:|-------:|-----------:|
132
+ | CC BY-SA 4.0 | 166187 | 41.388233 |
133
+ | CC0 | 112871 | 28.110088 |
134
+ | CC BY 4.0 | 112407 | 27.994531 |
135
+ | CC BY-NC-SA 4.0 | 4173 | 1.03927 |
136
+ | ODC-BY | 3769 | 0.938655 |
137
+ | CC BY-NC 4.0 | 1853 | 0.461483 |
138
+ | CC BY-NC-ND 4.0 | 250 | 0.062262 |
139
+ | CC BY 3.0 | 13 | 0.003238 |
140
+ | GFDL | 6 | 0.001494 |
141
+ | OGL | 3 | 0.000747 |