matejklemen commited on
Commit
6f5aaa3
1 Parent(s): 61a3f98

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +115 -0
README.md CHANGED
@@ -1,3 +1,118 @@
1
  ---
2
  license: cc-by-nc-sa-4.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-nc-sa-4.0
3
  ---
4
+
5
+ # Dataset Card for ccGigafida
6
+
7
+ This repository by default loads the publicly available dataset ccGigafida, which contains a small subset of the Gigafida/Gigafida2 corpus.
8
+ The full datasets are private due to copyright. **If you happen to have access to the full datasets, the script will also work with those.**
9
+ Instead of `datasets.load_dataset("cjvt/cc_gigafida")` please use `datasets.load_dataset("cjvt/cc_gigafida", data_dir="<directory-containing-gigafida(2)-TEI-files>", ignore_verifications=True)`.
10
+ **IMPORTANT:** The script will process all `.xml` files in the provided directory and its subdirectories - make sure there are no schema or metadata files in there!
11
+
12
+
13
+ ### Dataset Summary
14
+
15
+ ccGigafida is a reference corpus of Slovene texts. It is a publicly available subsample of an even larger reference corpus, Gigafida (and its successor Gigafida 2).
16
+ The Gigafida corpus is an extensive collection of Slovene text of various genres, from daily newspapers, magazines, all kinds of books (fiction, non-fiction, textbooks),
17
+ web pages, transcriptions of parliamentary debates and similar.
18
+
19
+ ### Supported Tasks and Leaderboards
20
+
21
+ Language modeling.
22
+
23
+ ### Languages
24
+
25
+ Slovenian.
26
+
27
+ ## Dataset Structure
28
+
29
+ ### Data Instances
30
+
31
+ The data is loaded at document-level, i.e. one instance is one document.
32
+ ```
33
+ {
34
+ 'id_doc': 'F0000123',
35
+ 'doc_title': 'Novi tednik NT&RC',
36
+ 'authors': ['neznani novinar'],
37
+ 'publish_date': '1998-03-27',
38
+ 'publisher': 'Novi tednik',
39
+ 'genres': ['tisk/periodično/časopis'],
40
+ 'doc_tokenized': [
41
+ [
42
+ ['Po', 'nekajletnem', 'počitku', 'pa', 'se', 'vračajo', 'tudi', 'kralji', 'dark', 'rock', 'godbe', 'JESUS', 'AND', 'THE', 'MARY', 'CHAIN', '.'],
43
+ ['Brata', 'Reid', 'bosta', 'svojo', 'najnovejšo', 'kreacijo', '»', 'Cracking', 'Up', '«', 'objavila', 'v', 'ponedeljek', 'pri', 'trenutno', 'najuspešnejši', 'neodvisni', 'založbi', 'Creation', '(', 'vodi', 'jo', 'njun', 'nekdanji', 'menager', 'Alan', 'McGee', ',', 'zanjo', 'pa', 'poleg', 'Oasis', 'snema', 'še', 'cel', 'kup', 'popularnih', 'brit', '-', 'popovcev', ')', ',', 'tej', 'pa', 'bo', 'kmalu', 'sledil', 'tudi', 'album', '»', 'Munki', '«', '.']
44
+ ],
45
+ [
46
+ ['Kultni', 'ameriški', 'tehno', 'freak', 'PLASTIKMAN', 'že', 'vrsto', 'let', 'velja', 'za', 'enega', 'izmed', 'najbolj', 'inovativnih', 'in', 'produktivnih', 'ustvarjalcev', 'sodobne', 'elektronske', 'glasbe', '.'],
47
+ ['Za', 'založbo', 'Nova', 'Mute', 'je', 'v', 'preteklih', 'nekaj', 'letih', 'posnel', 'cel', 'kup', 'izvrstnih', 'underground', 'dance', 'glasbenih', 'izdelkov', ',', 'pred', 'nedavnim', 'pa', 'je', 'ljubitelje', 'tovrstne', 'godbe', 'presenetil', 'z', 'ambientalnimi', 'odisejadami', ',', 'zbranimi', 'na', 'LP-ju', '»', 'Refused', '«', ',', 'ki', 'ga', 'lahko', 'od', 'prejšnjega', 'ponedeljka', 'kupite', 'tudi', 'v', 'bolje', 'založenih', 'trgovinah', 'z', 'nosilci', 'zvoka', 'na', 'sončni', 'strani', 'Alp', '.']
48
+ ],
49
+ [
50
+ ['STANE', 'ŠPEGEL']
51
+ ]
52
+ ],
53
+ 'doc_string': [
54
+ [
55
+ 'Po nekajletnem počitku pa se vračajo tudi kralji dark rock godbe JESUS AND THE MARY CHAIN. ',
56
+ 'Brata Reid bosta svojo najnovejšo kreacijo »Cracking Up« objavila v ponedeljek pri trenutno najuspešnejši neodvisni založbi Creation (vodi jo njun nekdanji menager Alan McGee, zanjo pa poleg Oasis snema še cel kup popularnih brit-popovcev), tej pa bo kmalu sledil tudi album »Munki«.'
57
+ ],
58
+ [
59
+ 'Kultni ameriški tehno freak PLASTIKMAN že vrsto let velja za enega izmed najbolj inovativnih in produktivnih ustvarjalcev sodobne elektronske glasbe. ',
60
+ 'Za založbo Nova Mute je v preteklih nekaj letih posnel cel kup izvrstnih underground dance glasbenih izdelkov, pred nedavnim pa je ljubitelje tovrstne godbe presenetil z ambientalnimi odisejadami, zbranimi na LP-ju »Refused«, ki ga lahko od prejšnjega ponedeljka kupite tudi v bolje založenih trgovinah z nosilci zvoka na sončni strani Alp.'
61
+ ],
62
+ [
63
+ 'STANE ŠPEGEL'
64
+ ]
65
+ ],
66
+ 'id_sents': [['F0000123.000005.0', 'F0000123.000005.1'], ['F0000123.000013.0', 'F0000123.000013.1'], ['F0000123.000020.0']]
67
+ }
68
+ ```
69
+
70
+
71
+ ### Data Fields
72
+
73
+ - `id_doc`: the document ID (string);
74
+ - `doc_title`: the document title (string);
75
+ - `authors`: author(s) of the document (list of string): "neznani novinar" (sl) = ("unknown/unspecified journalist");
76
+ - `publish_date`: publish date (string);
77
+ - `publisher`: publisher, e.g., the name of a news agency (string);
78
+ - `genres`: genre(s) of the document (list of string) - possible genres: `['tisk', 'tisk/knjižno', 'tisk/knjižno/leposlovno', 'tisk/knjižno/strokovno', 'tisk/periodično', 'tisk/periodično/časopis', 'tisk/periodično/revija', 'tisk/drugo', 'internet']`;
79
+ - `doc_tokenized`: tokenized document - the top level lists represent paragraphs, the lists in the level deeper represent sentences, and each sentence contains tokens;
80
+ - `doc_string`: same as `doc_tokenized` but with properly placed spaces in sentences;
81
+ - `id_sents`: IDs of sentences contained inside paragraphs of the document.
82
+
83
+ ## Dataset Creation
84
+
85
+ Gigafida consists of texts which were published between 1990 and 2011. The texts come from printed sources and from the web.
86
+ Printed part contains fiction, non-fiction and textbooks, and periodicals such as daily newspapers and magazines.
87
+ Texts originating from the web were published on news portals, pages of big Slovene companies and more important governmental,
88
+ educational, research, cultural and similar institutions.
89
+
90
+ For more information, please check http://eng.slovenscina.eu/korpusi/gigafida.
91
+
92
+ ## Additional Information
93
+
94
+ ### Dataset Curators
95
+
96
+ Nataša Logar; et al. (please see http://hdl.handle.net/11356/1035 for the full list)
97
+
98
+ ### Licensing Information
99
+
100
+ CC BY-NC-SA 4.0.
101
+
102
+ ### Citation Information
103
+
104
+ ```
105
+ @misc{ccGigafida,
106
+ title = {Written corpus {ccGigafida} 1.0},
107
+ author = {Logar, Nata{\v s}a and Erjavec, Toma{\v z} and Krek, Simon and Gr{\v c}ar, Miha and Holozan, Peter},
108
+ url = {http://hdl.handle.net/11356/1035},
109
+ note = {Slovenian language resource repository {CLARIN}.{SI}},
110
+ copyright = {Creative Commons - Attribution-{NonCommercial}-{ShareAlike} 4.0 International ({CC} {BY}-{NC}-{SA} 4.0)},
111
+ issn = {2820-4042},
112
+ year = {2013}
113
+ }
114
+ ```
115
+
116
+ ### Contributions
117
+
118
+ Thanks to [@matejklemen](https://github.com/matejklemen) for adding this dataset.