Matej Klemen commited on
Commit
8adda2c
1 Parent(s): 386ed04

Add first version of RSDO4 en-sl corpus

Browse files
Files changed (1) hide show
  1. rsdo4_en_sl.py +85 -0
rsdo4_en_sl.py ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """ rsdo4_en_sl is a parallel corpus of English-Slovene and Slovene-English translation pairs."""
2
+ import xml.etree.ElementTree as ET
3
+ import os
4
+
5
+ import datasets
6
+
7
+
8
+ _CITATION = """\
9
+ @misc{rsdo4_en_sl,
10
+ title = {Parallel corpus {EN}-{SL} {RSDO4} 1.0},
11
+ author = {Repar, Andra{\v z} and Lebar Bajec, Iztok},
12
+ url = {http://hdl.handle.net/11356/1457},
13
+ year = {2021}
14
+ }
15
+ """
16
+
17
+ _DESCRIPTION = """\
18
+ The RSDO4 parallel corpus of English-Slovene and Slovene-English translation pairs was collected as part of work
19
+ package 4 of the Slovene in the Digital Environment project. It contains texts collected from public institutions
20
+ and texts submitted by individual donors through the text collection portal created within the project. The corpus
21
+ consists of 964433 translation pairs (extracted from standard translation formats (TMX, XLIFF) or manually aligned)
22
+ in randomized order which can be used for machine translation training.
23
+ """
24
+
25
+ _HOMEPAGE = "http://hdl.handle.net/11356/1457"
26
+
27
+ _LICENSE = "Creative Commons - Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)"
28
+
29
+ _URLS = {
30
+ "rsdo4_en_sl": "https://www.clarin.si/repository/xmlui/bitstream/handle/11356/1457/rsdo4-parallel-corpus-1-0.tmx.zip"
31
+ }
32
+
33
+ XML_NAMESPACE = "{http://www.w3.org/XML/1998/namespace}"
34
+
35
+
36
+ class RSDO4EnSl(datasets.GeneratorBasedBuilder):
37
+ """rsdo4_en_sl is a parallel corpus of English-Slovene and Slovene-English translation pairs."""
38
+
39
+ VERSION = datasets.Version("1.0.0")
40
+
41
+ def _info(self):
42
+ features = datasets.Features(
43
+ {
44
+ "en_seq": datasets.Value("string"),
45
+ "sl_seq": datasets.Value("string"),
46
+ }
47
+ )
48
+ return datasets.DatasetInfo(
49
+ description=_DESCRIPTION,
50
+ features=features,
51
+ homepage=_HOMEPAGE,
52
+ license=_LICENSE,
53
+ citation=_CITATION,
54
+ )
55
+
56
+ def _split_generators(self, dl_manager):
57
+ urls = _URLS["rsdo4_en_sl"]
58
+ data_dir = dl_manager.download_and_extract(urls)
59
+ return [
60
+ datasets.SplitGenerator(
61
+ name=datasets.Split.TRAIN,
62
+ # These kwargs will be passed to _generate_examples
63
+ gen_kwargs={
64
+ "file_path": os.path.join(data_dir, "rsdo4-parallel-corpus-1-0.tmx")
65
+ }
66
+ )
67
+ ]
68
+
69
+ def _generate_examples(self, file_path):
70
+ doc = ET.parse(file_path)
71
+ root = doc.getroot()
72
+
73
+ for i, curr_unit in enumerate(root.iterfind(".//tu")):
74
+ translation_pair = curr_unit.findall("tuv")
75
+
76
+ formatted_pair = {}
77
+ for curr_seq in translation_pair:
78
+ # Not sure whether the languages are erroneously swapped or whether xml:lang represents the "language direction"
79
+ seq_lang = "sl" if curr_seq.attrib[f"{XML_NAMESPACE}lang"] == "en" else "en"
80
+
81
+ curr_text = curr_seq.find("seg").text
82
+ curr_text = "" if curr_text is None else curr_text.strip()
83
+ formatted_pair[f"{seq_lang}_seq"] = curr_text
84
+
85
+ yield i, formatted_pair