Remove duplicate claims
#1
by
azhx
- opened
fever.py
CHANGED
|
@@ -46,7 +46,6 @@ class FeverConfig(datasets.BuilderConfig):
|
|
| 46 |
|
| 47 |
class Fever(datasets.GeneratorBasedBuilder):
|
| 48 |
"""Fact Extraction and VERification Dataset."""
|
| 49 |
-
|
| 50 |
BUILDER_CONFIGS = [
|
| 51 |
FeverConfig(
|
| 52 |
name="v1.0",
|
|
@@ -134,14 +133,21 @@ class Fever(datasets.GeneratorBasedBuilder):
|
|
| 134 |
),
|
| 135 |
]
|
| 136 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
def _info(self):
|
| 138 |
-
if
|
|
|
|
| 139 |
features = {
|
| 140 |
"id": datasets.Value("string"),
|
| 141 |
"text": datasets.Value("string"),
|
| 142 |
"lines": datasets.Value("string"),
|
| 143 |
}
|
| 144 |
-
elif
|
|
|
|
|
|
|
| 145 |
features = {
|
| 146 |
"id": datasets.Value("int32"),
|
| 147 |
"label": datasets.ClassLabel(names=["REFUTES", "SUPPORTS"]),
|
|
@@ -152,10 +158,13 @@ class Fever(datasets.GeneratorBasedBuilder):
|
|
| 152 |
"evidence_sentence_id": datasets.Value("int32"),
|
| 153 |
}
|
| 154 |
return datasets.DatasetInfo(
|
| 155 |
-
description=
|
|
|
|
| 156 |
features=datasets.Features(features),
|
| 157 |
-
homepage=
|
| 158 |
-
|
|
|
|
|
|
|
| 159 |
)
|
| 160 |
|
| 161 |
def _split_generators(self, dl_manager):
|
|
@@ -187,6 +196,10 @@ class Fever(datasets.GeneratorBasedBuilder):
|
|
| 187 |
continue
|
| 188 |
|
| 189 |
claim = data["claim"]
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
evidences = data.get("evidence", [])
|
| 191 |
if len(evidences) > 0:
|
| 192 |
for i in range(len(evidences)):
|
|
|
|
| 46 |
|
| 47 |
class Fever(datasets.GeneratorBasedBuilder):
|
| 48 |
"""Fact Extraction and VERification Dataset."""
|
|
|
|
| 49 |
BUILDER_CONFIGS = [
|
| 50 |
FeverConfig(
|
| 51 |
name="v1.0",
|
|
|
|
| 133 |
),
|
| 134 |
]
|
| 135 |
|
| 136 |
+
def __init__(self, *args, **kwargs):
|
| 137 |
+
super().__init__(*args, **kwargs)
|
| 138 |
+
self.seen_claims= set()
|
| 139 |
+
|
| 140 |
def _info(self):
|
| 141 |
+
if
|
| 142 |
+
config.name == "wiki_pages":
|
| 143 |
features = {
|
| 144 |
"id": datasets.Value("string"),
|
| 145 |
"text": datasets.Value("string"),
|
| 146 |
"lines": datasets.Value("string"),
|
| 147 |
}
|
| 148 |
+
elif
|
| 149 |
+
config.name == "v1.0" or
|
| 150 |
+
config.name == "v2.0":
|
| 151 |
features = {
|
| 152 |
"id": datasets.Value("int32"),
|
| 153 |
"label": datasets.ClassLabel(names=["REFUTES", "SUPPORTS"]),
|
|
|
|
| 158 |
"evidence_sentence_id": datasets.Value("int32"),
|
| 159 |
}
|
| 160 |
return datasets.DatasetInfo(
|
| 161 |
+
description=
|
| 162 |
+
config.description,
|
| 163 |
features=datasets.Features(features),
|
| 164 |
+
homepage=
|
| 165 |
+
config.homepage,
|
| 166 |
+
citation=
|
| 167 |
+
config.citation,
|
| 168 |
)
|
| 169 |
|
| 170 |
def _split_generators(self, dl_manager):
|
|
|
|
| 196 |
continue
|
| 197 |
|
| 198 |
claim = data["claim"]
|
| 199 |
+
if claim in self.seen_claims:
|
| 200 |
+
continue
|
| 201 |
+
else:
|
| 202 |
+
self.seen_claims.add(claim)
|
| 203 |
evidences = data.get("evidence", [])
|
| 204 |
if len(evidences) > 0:
|
| 205 |
for i in range(len(evidences)):
|