holylovenia
commited on
Commit
·
351a67d
1
Parent(s):
33e265f
Upload talpco.py with huggingface_hub
Browse files
talpco.py
ADDED
@@ -0,0 +1,160 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from pathlib import Path
|
3 |
+
from typing import Dict, List
|
4 |
+
|
5 |
+
import datasets
|
6 |
+
|
7 |
+
from nusacrowd.utils import schemas
|
8 |
+
from nusacrowd.utils.configs import NusantaraConfig
|
9 |
+
from nusacrowd.utils.constants import Tasks
|
10 |
+
|
11 |
+
_CITATION = """\
|
12 |
+
@article{published_papers/22434604,
|
13 |
+
title = {TUFS Asian Language Parallel Corpus (TALPCo)},
|
14 |
+
author = {Hiroki Nomoto and Kenji Okano and David Moeljadi and Hideo Sawada},
|
15 |
+
journal = {言語処理学会 第24回年次大会 発表論文集},
|
16 |
+
pages = {436--439},
|
17 |
+
year = {2018}
|
18 |
+
}
|
19 |
+
@article{published_papers/22434603,
|
20 |
+
title = {Interpersonal meaning annotation for Asian language corpora: The case of TUFS Asian Language Parallel Corpus (TALPCo)},
|
21 |
+
author = {Hiroki Nomoto and Kenji Okano and Sunisa Wittayapanyanon and Junta Nomura},
|
22 |
+
journal = {言語処理学会 第25回年次大会 発表論文集},
|
23 |
+
pages = {846--849},
|
24 |
+
year = {2019}
|
25 |
+
}
|
26 |
+
"""
|
27 |
+
_DATASETNAME = "talpco"
|
28 |
+
_DESCRIPTION = """\
|
29 |
+
The TUFS Asian Language Parallel Corpus (TALPCo) is an open parallel corpus consisting of Japanese sentences
|
30 |
+
and their translations into Korean, Burmese (Myanmar; the official language of the Republic of the Union of Myanmar),
|
31 |
+
Malay (the national language of Malaysia, Singapore and Brunei), Indonesian, Thai, Vietnamese and English.
|
32 |
+
"""
|
33 |
+
_HOMEPAGE = "https://github.com/matbahasa/TALPCo"
|
34 |
+
_LOCAL = False
|
35 |
+
_LANGUAGES = ["eng", "ind", "jpn", "kor", "myn", "tha", "vie", "zsm"]
|
36 |
+
_LICENSE = "CC-BY 4.0"
|
37 |
+
_URLS = {
|
38 |
+
_DATASETNAME: "https://github.com/matbahasa/TALPCo/archive/refs/heads/master.zip",
|
39 |
+
}
|
40 |
+
_SUPPORTED_TASKS = [Tasks.MACHINE_TRANSLATION]
|
41 |
+
_SOURCE_VERSION = "1.0.0"
|
42 |
+
_NUSANTARA_VERSION = "1.0.0"
|
43 |
+
|
44 |
+
|
45 |
+
def nusantara_config_constructor(lang_source, lang_target, schema, version):
|
46 |
+
"""Construct NusantaraConfig with talpco_{lang_source}_{lang_target}_{schema} as the name format"""
|
47 |
+
if schema != "source" and schema != "nusantara_t2t":
|
48 |
+
raise ValueError(f"Invalid schema: {schema}")
|
49 |
+
|
50 |
+
if lang_source == "" and lang_target == "":
|
51 |
+
return NusantaraConfig(
|
52 |
+
name="talpco_{schema}".format(schema=schema),
|
53 |
+
version=datasets.Version(version),
|
54 |
+
description="talpco with {schema} schema for all 7 language pairs from / to ind language".format(schema=schema),
|
55 |
+
schema=schema,
|
56 |
+
subset_id="talpco",
|
57 |
+
)
|
58 |
+
else:
|
59 |
+
return NusantaraConfig(
|
60 |
+
name="talpco_{lang_source}_{lang_target}_{schema}".format(lang_source=lang_source, lang_target=lang_target, schema=schema),
|
61 |
+
version=datasets.Version(version),
|
62 |
+
description="talpco with {schema} schema for {lang_source} source language and {lang_target} target language".format(lang_source=lang_source, lang_target=lang_target, schema=schema),
|
63 |
+
schema=schema,
|
64 |
+
subset_id="talpco",
|
65 |
+
)
|
66 |
+
|
67 |
+
|
68 |
+
class TALPCo(datasets.GeneratorBasedBuilder):
|
69 |
+
"""TALPCo datasets contains 1372 datasets in 8 languages"""
|
70 |
+
|
71 |
+
BUILDER_CONFIGS = (
|
72 |
+
[nusantara_config_constructor(lang1, lang2, "source", _SOURCE_VERSION) for lang1 in _LANGUAGES for lang2 in _LANGUAGES if lang1 != lang2]
|
73 |
+
+ [nusantara_config_constructor(lang1, lang2, "nusantara_t2t", _NUSANTARA_VERSION) for lang1 in _LANGUAGES for lang2 in _LANGUAGES if lang1 != lang2]
|
74 |
+
+ [nusantara_config_constructor("", "", "source", _SOURCE_VERSION), nusantara_config_constructor("", "", "nusantara_t2t", _NUSANTARA_VERSION)]
|
75 |
+
)
|
76 |
+
|
77 |
+
DEFAULT_CONFIG_NAME = "talpco_jpn_ind_source"
|
78 |
+
|
79 |
+
def _info(self) -> datasets.DatasetInfo:
|
80 |
+
if self.config.schema == "source" or self.config.schema == "nusantara_t2t":
|
81 |
+
features = schemas.text2text_features
|
82 |
+
else:
|
83 |
+
raise ValueError(f"Invalid config schema: {self.config.schema}")
|
84 |
+
|
85 |
+
return datasets.DatasetInfo(
|
86 |
+
description=_DESCRIPTION,
|
87 |
+
features=features,
|
88 |
+
homepage=_HOMEPAGE,
|
89 |
+
license=_LICENSE,
|
90 |
+
citation=_CITATION,
|
91 |
+
)
|
92 |
+
|
93 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
94 |
+
urls = _URLS[_DATASETNAME]
|
95 |
+
base_path = Path(dl_manager.download_and_extract(urls)) / "TALPCo-master"
|
96 |
+
data = {}
|
97 |
+
for lang in _LANGUAGES:
|
98 |
+
lang_file_name = "data_" + lang + ".txt"
|
99 |
+
lang_file_path = base_path / lang / lang_file_name
|
100 |
+
if os.path.isfile(lang_file_path):
|
101 |
+
with open(lang_file_path, "r") as file:
|
102 |
+
data[lang] = file.read().strip("\n").split("\n")
|
103 |
+
|
104 |
+
return [
|
105 |
+
datasets.SplitGenerator(
|
106 |
+
name=datasets.Split.TRAIN,
|
107 |
+
gen_kwargs={
|
108 |
+
"data": data,
|
109 |
+
"split": "train",
|
110 |
+
},
|
111 |
+
),
|
112 |
+
]
|
113 |
+
|
114 |
+
def _generate_examples(self, data: Dict, split: str):
|
115 |
+
if self.config.schema != "source" and self.config.schema != "nusantara_t2t":
|
116 |
+
raise ValueError(f"Invalid config schema: {self.config.schema}")
|
117 |
+
|
118 |
+
if self.config.name == "talpco_source" or self.config.name == "talpco_nusantara_t2t":
|
119 |
+
# load all 7 language pairs from / to ind language
|
120 |
+
lang_target = "ind"
|
121 |
+
for lang_source in _LANGUAGES:
|
122 |
+
if lang_source == lang_target:
|
123 |
+
continue
|
124 |
+
for language_pair_data in self.generate_language_pair_data(lang_source, lang_target, data):
|
125 |
+
yield language_pair_data
|
126 |
+
|
127 |
+
lang_source = "ind"
|
128 |
+
for lang_target in _LANGUAGES:
|
129 |
+
if lang_source == lang_target:
|
130 |
+
continue
|
131 |
+
for language_pair_data in self.generate_language_pair_data(lang_source, lang_target, data):
|
132 |
+
yield language_pair_data
|
133 |
+
else:
|
134 |
+
_, lang_source, lang_target = self.config.name.replace(f"_{self.config.schema}", "").split("_")
|
135 |
+
for language_pair_data in self.generate_language_pair_data(lang_source, lang_target, data):
|
136 |
+
yield language_pair_data
|
137 |
+
|
138 |
+
def generate_language_pair_data(self, lang_source, lang_target, data):
|
139 |
+
dict_source = {}
|
140 |
+
for row in data[lang_source]:
|
141 |
+
id, text = row.split("\t")
|
142 |
+
dict_source[id] = text
|
143 |
+
|
144 |
+
dict_target = {}
|
145 |
+
for row in data[lang_target]:
|
146 |
+
id, text = row.split("\t")
|
147 |
+
dict_target[id] = text
|
148 |
+
|
149 |
+
all_ids = set([k for k in dict_source.keys()] + [k for k in dict_target.keys()])
|
150 |
+
dict_merged = {k: [dict_source.get(k), dict_target.get(k)] for k in all_ids}
|
151 |
+
|
152 |
+
for id in sorted(all_ids):
|
153 |
+
ex = {
|
154 |
+
"id": lang_source + "_" + lang_target + "_" + id,
|
155 |
+
"text_1": dict_merged[id][0],
|
156 |
+
"text_2": dict_merged[id][1],
|
157 |
+
"text_1_name": lang_source,
|
158 |
+
"text_2_name": lang_target,
|
159 |
+
}
|
160 |
+
yield lang_source + "_" + lang_target + "_" + id, ex
|