Upload lbox_open.py
Browse files- lbox_open.py +142 -0
lbox_open.py
ADDED
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# LBox
|
2 |
+
import json
|
3 |
+
import datasets
|
4 |
+
|
5 |
+
_CASENAME_CLASSIFICATION_FEATURES = {
|
6 |
+
"id": datasets.Value("int32"),
|
7 |
+
"casetype": datasets.Value("string"),
|
8 |
+
"casename": datasets.Value("string"),
|
9 |
+
"facts": datasets.Value("string"),
|
10 |
+
}
|
11 |
+
|
12 |
+
_STATUTE_CLASSIFICATION_FEATURES = {
|
13 |
+
"id": datasets.Value("int32"),
|
14 |
+
"casetype": datasets.Value("string"),
|
15 |
+
"casename": datasets.Value("string"),
|
16 |
+
"statutes": datasets.features.Sequence(datasets.Value("string")),
|
17 |
+
"facts": datasets.Value("string"),
|
18 |
+
}
|
19 |
+
|
20 |
+
_SUMMARIZATION_CLASSIFICATION_FEATURES = {
|
21 |
+
"id": datasets.Value("int32"),
|
22 |
+
"summary": datasets.Value("string"),
|
23 |
+
"precedent": datasets.Value("string"),
|
24 |
+
}
|
25 |
+
|
26 |
+
_CASE_CORPUS_FEATURES = {
|
27 |
+
"id": datasets.Value("int32"),
|
28 |
+
"precedent": datasets.Value("string"),
|
29 |
+
}
|
30 |
+
|
31 |
+
|
32 |
+
class LBoxOpenConfig(datasets.BuilderConfig):
|
33 |
+
"""BuilderConfig for OpenLBox."""
|
34 |
+
|
35 |
+
def __init__(self, features, data_url, citation, url, label_classes=("False", "True"), **kwargs):
|
36 |
+
# Version history:
|
37 |
+
# 0.1.0: Initial version.
|
38 |
+
super(LBoxOpenConfig, self).__init__(version=datasets.Version("0.1.0"), **kwargs)
|
39 |
+
self.features = features
|
40 |
+
self.label_classes = label_classes
|
41 |
+
self.data_url = data_url
|
42 |
+
self.citation = citation
|
43 |
+
self.url = url
|
44 |
+
|
45 |
+
|
46 |
+
class LBoxOpen(datasets.GeneratorBasedBuilder):
|
47 |
+
"""The Legal AI Benchmark dataset from Korean Legal Cases."""
|
48 |
+
BUILDER_CONFIGS = [
|
49 |
+
LBoxOpenConfig(
|
50 |
+
name="casename_classification",
|
51 |
+
description="",
|
52 |
+
features=_CASENAME_CLASSIFICATION_FEATURES,
|
53 |
+
data_url="https://lbox-open.s3.ap-northeast-2.amazonaws.com/precedent_benchmark_dataset/casename_classification/v0.1.0/",
|
54 |
+
citation="",
|
55 |
+
url="lbox.kr",
|
56 |
+
),
|
57 |
+
LBoxOpenConfig(
|
58 |
+
name="statute_classification",
|
59 |
+
description="",
|
60 |
+
features=_STATUTE_CLASSIFICATION_FEATURES,
|
61 |
+
data_url="https://lbox-open.s3.ap-northeast-2.amazonaws.com/precedent_benchmark_dataset/statute_classification/v0.1.0/",
|
62 |
+
citation="",
|
63 |
+
url="lbox.kr",
|
64 |
+
),
|
65 |
+
LBoxOpenConfig(
|
66 |
+
name="summarization",
|
67 |
+
description="",
|
68 |
+
features=_SUMMARIZATION_CLASSIFICATION_FEATURES,
|
69 |
+
data_url="https://lbox-open.s3.ap-northeast-2.amazonaws.com/precedent_benchmark_dataset/summarization/v0.1.0/",
|
70 |
+
citation="",
|
71 |
+
url="lbox.kr",
|
72 |
+
),
|
73 |
+
LBoxOpenConfig(
|
74 |
+
name="case_corpus",
|
75 |
+
description="",
|
76 |
+
features=_CASE_CORPUS_FEATURES,
|
77 |
+
data_url="https://lbox-open.s3.ap-northeast-2.amazonaws.com/precedent_benchmark_dataset/case_corpus/v0.1.0/",
|
78 |
+
citation="",
|
79 |
+
url="lbox.kr",
|
80 |
+
),
|
81 |
+
]
|
82 |
+
|
83 |
+
def _info(self):
|
84 |
+
return datasets.DatasetInfo(
|
85 |
+
description="",
|
86 |
+
features=datasets.Features(self.config.features),
|
87 |
+
homepage=self.config.url,
|
88 |
+
citation="",
|
89 |
+
)
|
90 |
+
|
91 |
+
def _split_generators(self, dl_manager):
|
92 |
+
if self.config.name == "case_corpus":
|
93 |
+
dl_dir = {
|
94 |
+
"train": dl_manager.download_and_extract(f"{self.config.data_url}case_corpus-150k.jsonl") or "",
|
95 |
+
}
|
96 |
+
|
97 |
+
return [
|
98 |
+
datasets.SplitGenerator(
|
99 |
+
name=datasets.Split.TRAIN,
|
100 |
+
gen_kwargs={
|
101 |
+
"data_file": dl_dir["train"],
|
102 |
+
"split": datasets.Split.TRAIN,
|
103 |
+
},
|
104 |
+
)
|
105 |
+
]
|
106 |
+
|
107 |
+
else:
|
108 |
+
dl_dir = {
|
109 |
+
"train": dl_manager.download_and_extract(f"{self.config.data_url}train.jsonl") or "",
|
110 |
+
"valid": dl_manager.download_and_extract(f"{self.config.data_url}valid.jsonl") or "",
|
111 |
+
"test": dl_manager.download_and_extract(f"{self.config.data_url}test.jsonl") or "",
|
112 |
+
}
|
113 |
+
|
114 |
+
return [
|
115 |
+
datasets.SplitGenerator(
|
116 |
+
name=datasets.Split.TRAIN,
|
117 |
+
gen_kwargs={
|
118 |
+
"data_file": dl_dir["train"],
|
119 |
+
"split": datasets.Split.TRAIN,
|
120 |
+
},
|
121 |
+
),
|
122 |
+
datasets.SplitGenerator(
|
123 |
+
name=datasets.Split.VALIDATION,
|
124 |
+
gen_kwargs={
|
125 |
+
"data_file": dl_dir["valid"],
|
126 |
+
"split": datasets.Split.VALIDATION,
|
127 |
+
},
|
128 |
+
),
|
129 |
+
datasets.SplitGenerator(
|
130 |
+
name=datasets.Split.TEST,
|
131 |
+
gen_kwargs={
|
132 |
+
"data_file": dl_dir["test"],
|
133 |
+
"split": datasets.Split.TEST,
|
134 |
+
},
|
135 |
+
),
|
136 |
+
]
|
137 |
+
|
138 |
+
def _generate_examples(self, data_file, split):
|
139 |
+
with open(data_file, encoding="utf-8") as f:
|
140 |
+
for line in f:
|
141 |
+
row = json.loads(line)
|
142 |
+
yield row["id"], row
|