Upload selfkgdwy100kdbpyg.py
Browse files- selfkgdwy100kdbpyg.py +203 -0
selfkgdwy100kdbpyg.py
ADDED
@@ -0,0 +1,203 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import os
|
3 |
+
import pickle
|
4 |
+
|
5 |
+
import datasets
|
6 |
+
|
7 |
+
logger = datasets.logging.get_logger(__name__)
|
8 |
+
|
9 |
+
_SUBFIELD = "yg"
|
10 |
+
|
11 |
+
_VERSION = "1.0.0"
|
12 |
+
|
13 |
+
_DESCRIPTION = """\
|
14 |
+
DWY100k-yg is a large-scale monolingual dataset extracted from DBpedia and YAGO3. The suffix yg means DBpedia
|
15 |
+
to YAGO3. And DWY100k-yg has 100,000 reference entity alignments.
|
16 |
+
"""
|
17 |
+
|
18 |
+
_CITATION = """\
|
19 |
+
@inproceedings{sun2018bootstrapping,
|
20 |
+
title={Bootstrapping Entity Alignment with Knowledge Graph Embedding.},
|
21 |
+
author={Sun, Zequn and Hu, Wei and Zhang, Qingheng and Qu, Yuzhong},
|
22 |
+
booktitle={IJCAI},
|
23 |
+
volume={18},
|
24 |
+
pages={4396--4402},
|
25 |
+
year={2018}
|
26 |
+
}
|
27 |
+
"""
|
28 |
+
|
29 |
+
_URL = "https://dl.acm.org/doi/10.1145/3485447.3511945"
|
30 |
+
|
31 |
+
_PREFIX = "https://huggingface.co/datasets/matchbench/selfkg-dwy100k-dbpyg"
|
32 |
+
|
33 |
+
_URLS = {
|
34 |
+
"source": f"{_PREFIX}/resolve/main/selfkg-dwy100k-dbp{_SUBFIELD}-src.zip",
|
35 |
+
"target": f"{_PREFIX}/resolve/main/selfkg-dwy100k-dbp{_SUBFIELD}-tgt.zip",
|
36 |
+
"pairs": f"{_PREFIX}/resolve/main/selfkg-dwy100k-dbp{_SUBFIELD}-pairs.zip",
|
37 |
+
}
|
38 |
+
|
39 |
+
class SelfkgDwy100kygConfig(datasets.BuilderConfig):
|
40 |
+
"""BuilderConfig for Selfkg-DWY100k."""
|
41 |
+
|
42 |
+
def __init__(self, features, data_url, citation, url, label_classes=("False", "True"), **kwargs):
|
43 |
+
"""
|
44 |
+
Args:
|
45 |
+
**kwargs: keyword arguments forwarded to super.
|
46 |
+
"""
|
47 |
+
super(SelfkgDwy100kygConfig, self).__init__(**kwargs)
|
48 |
+
self.features = features
|
49 |
+
self.label_classes = label_classes
|
50 |
+
self.data_url = data_url
|
51 |
+
self.citation = citation
|
52 |
+
self.url = url
|
53 |
+
|
54 |
+
class DWY100kYg(datasets.GeneratorBasedBuilder):
|
55 |
+
"""DWY100k-yg: A Entity Alignment Dataset. From DBpedia to YAGO3."""
|
56 |
+
|
57 |
+
BUILDER_CONFIGS = [
|
58 |
+
SelfkgDwy100kygConfig(
|
59 |
+
name="source",
|
60 |
+
features=["column1", "column2", "column3"],
|
61 |
+
citation="TODO",
|
62 |
+
url="TODO",
|
63 |
+
data_url="https://huggingface.co/datasets/matchbench/selfkg-dwy100k-dbpyg/resolve/main/selfkg-dwy100k-dbpyg.zip"
|
64 |
+
),
|
65 |
+
SelfkgDwy100kygConfig(
|
66 |
+
name="target",
|
67 |
+
features=["column1", "column2", "column3"],
|
68 |
+
citation="TODO",
|
69 |
+
url="TODO",
|
70 |
+
data_url="https://huggingface.co/datasets/matchbench/selfkg-dwy100k-dbpyg/resolve/main/selfkg-dwy100k-dbpyg.zip"
|
71 |
+
),
|
72 |
+
SelfkgDwy100kygConfig(
|
73 |
+
name="pairs",
|
74 |
+
features=["left_id","right_id"],
|
75 |
+
citation="TODO",
|
76 |
+
url="TODO",
|
77 |
+
data_url="https://huggingface.co/datasets/matchbench/selfkg-dwy100k-dbpyg/resolve/main/selfkg-dwy100k-dbpyg.zip"
|
78 |
+
)
|
79 |
+
]
|
80 |
+
|
81 |
+
def _info(self) -> datasets.DatasetInfo:
|
82 |
+
if self.config.name=="source":
|
83 |
+
features = {feature: datasets.Value("string") for feature in self.config.features}
|
84 |
+
elif self.config.name=="target":
|
85 |
+
features = {feature: datasets.Value("string") for feature in self.config.features}
|
86 |
+
elif self.config.name=="pairs":
|
87 |
+
features = {feature: datasets.Value("int32") for feature in self.config.features}
|
88 |
+
|
89 |
+
return datasets.DatasetInfo(description = _DESCRIPTION,
|
90 |
+
features = datasets.Features(features))
|
91 |
+
|
92 |
+
|
93 |
+
def _split_generators(self, dl_manager):
|
94 |
+
|
95 |
+
dl_dir = dl_manager.download_and_extract(self.config.data_url) or ""
|
96 |
+
if self.config.name == "source":
|
97 |
+
return [
|
98 |
+
datasets.SplitGenerator(
|
99 |
+
name="ent_ids",
|
100 |
+
gen_kwargs={
|
101 |
+
"data_file": os.path.join(dl_dir, "id_ent_1"),
|
102 |
+
"split": "ent_ids",
|
103 |
+
},
|
104 |
+
),
|
105 |
+
datasets.SplitGenerator(
|
106 |
+
name="rel_triples_id",
|
107 |
+
gen_kwargs={
|
108 |
+
"data_file": os.path.join(dl_dir, "triples_1"),
|
109 |
+
"split": "rel_triples",
|
110 |
+
},
|
111 |
+
),
|
112 |
+
datasets.SplitGenerator(
|
113 |
+
name="LaBSE_emb",
|
114 |
+
gen_kwargs={
|
115 |
+
"data_file": os.path.join(dl_dir, "raw_LaBSE_emb_1.pkl"),
|
116 |
+
"split": "embedding",
|
117 |
+
},
|
118 |
+
),
|
119 |
+
]
|
120 |
+
elif self.config.name == "target":
|
121 |
+
return [
|
122 |
+
datasets.SplitGenerator(
|
123 |
+
name="ent_ids",
|
124 |
+
gen_kwargs={
|
125 |
+
"data_file": os.path.join(dl_dir, "id_ent_2"),
|
126 |
+
"split": "ent_ids",
|
127 |
+
},
|
128 |
+
),
|
129 |
+
datasets.SplitGenerator(
|
130 |
+
name="rel_triples_id",
|
131 |
+
gen_kwargs={
|
132 |
+
"data_file": os.path.join(dl_dir, "triples_2"),
|
133 |
+
"split": "rel_triples",
|
134 |
+
},
|
135 |
+
),
|
136 |
+
datasets.SplitGenerator(
|
137 |
+
name="LaBSE_emb",
|
138 |
+
gen_kwargs={
|
139 |
+
"data_file": os.path.join(dl_dir, "raw_LaBSE_emb_2.pkl"),
|
140 |
+
"split": "embedding",
|
141 |
+
},
|
142 |
+
),
|
143 |
+
]
|
144 |
+
elif self.config.name == "pairs":
|
145 |
+
return [
|
146 |
+
datasets.SplitGenerator(
|
147 |
+
name="train",
|
148 |
+
gen_kwargs={
|
149 |
+
"data_file": os.path.join(dl_dir, "ref_ent_ids"),
|
150 |
+
"split": "train",
|
151 |
+
},
|
152 |
+
),
|
153 |
+
datasets.SplitGenerator(
|
154 |
+
name="valid",
|
155 |
+
gen_kwargs={
|
156 |
+
"data_file": os.path.join(dl_dir, "valid.ref"),
|
157 |
+
"split": "valid",
|
158 |
+
},
|
159 |
+
),
|
160 |
+
datasets.SplitGenerator(
|
161 |
+
name="test",
|
162 |
+
gen_kwargs={
|
163 |
+
"data_file": os.path.join(dl_dir, "ref_ent_ids"),
|
164 |
+
"split": "test",
|
165 |
+
},
|
166 |
+
),
|
167 |
+
]
|
168 |
+
|
169 |
+
def _generate_examples(self, data_file, split):
|
170 |
+
if split in ["LaBSE_emb"]:
|
171 |
+
des = pickle.load(open(data_file,"rb"))
|
172 |
+
i = -1
|
173 |
+
for ent,ori_emb in des.items():
|
174 |
+
i += 1
|
175 |
+
yield i, {
|
176 |
+
"column1": ent,
|
177 |
+
"column2": ori_emb,
|
178 |
+
"column3": None
|
179 |
+
}
|
180 |
+
else:
|
181 |
+
f = open(data_file,"r", encoding='utf-8')
|
182 |
+
data = f.readlines()
|
183 |
+
for i in range(len(data)):
|
184 |
+
if self.config.name in ["source", "target"]:
|
185 |
+
if split in ["ent_ids", "rel_ids"]:
|
186 |
+
row = data[i].strip('\n').split('\t')
|
187 |
+
yield i, {
|
188 |
+
"id": row[0],
|
189 |
+
"data": row[1],
|
190 |
+
}
|
191 |
+
elif split in ["rel_triples_id"]:
|
192 |
+
row = data[i].strip('\n').split('\t')
|
193 |
+
yield i, {
|
194 |
+
"head_ent": row[0],
|
195 |
+
"relation": row[1],
|
196 |
+
"tail_ent": row[2]
|
197 |
+
}
|
198 |
+
if self.config.name == "pairs":
|
199 |
+
row = data[i].strip('\n').split('\t')
|
200 |
+
yield i, {
|
201 |
+
"left_id": row[0],
|
202 |
+
"right_id": row[1]
|
203 |
+
}
|