Datasets:
Update DiaBLa.py
Browse files
DiaBLa.py
CHANGED
|
@@ -50,9 +50,9 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 50 |
|
| 51 |
BUILDER_CONFIGS = [
|
| 52 |
DiablaConfig(
|
| 53 |
-
name=
|
| 54 |
-
version=datasets.Version(
|
| 55 |
-
description=
|
| 56 |
),
|
| 57 |
]
|
| 58 |
|
|
@@ -62,11 +62,11 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 62 |
description=_DESCRIPTION,
|
| 63 |
features=datasets.Features(
|
| 64 |
{
|
| 65 |
-
'id': datasets.Value(
|
| 66 |
-
'orig': datasets.Value(
|
| 67 |
-
'norm': datasets.Value(
|
| 68 |
-
'mt': datasets.Value(
|
| 69 |
-
'ref': datasets.Value(
|
| 70 |
'utterance_meta': datasets.features.Sequence(
|
| 71 |
{
|
| 72 |
'eval-judgment': ClassLabel(num_classes=3, names=['poor', 'medium', 'perfect']),
|
|
@@ -81,19 +81,27 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 81 |
),
|
| 82 |
'dialogue_meta': datasets.features.Sequence(
|
| 83 |
{
|
| 84 |
-
'start_time': datasets.Value(
|
| 85 |
-
'end_time' : datasets.Value(
|
| 86 |
-
'translation_model': datasets.Value(
|
| 87 |
'final_evaluation_user1': datasets.features.Sequence(
|
| 88 |
{
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
'final_evaluation_user2': datasets.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
'scenario': datasets.features.Sequence(
|
| 98 |
[
|
| 99 |
[
|
|
@@ -103,20 +111,20 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 103 |
),
|
| 104 |
'user1': datasets.features.Sequence(
|
| 105 |
{
|
| 106 |
-
'rolenum': datasets.Value(
|
| 107 |
-
'role': datasets.Value(
|
| 108 |
-
'initiated_dialogue': datasets.Value(
|
| 109 |
-
'turn_number': datasets.Value(
|
| 110 |
-
'lang': datasets.Value(
|
| 111 |
}
|
| 112 |
),
|
| 113 |
'user2': datasets.features.Sequence(
|
| 114 |
{
|
| 115 |
-
'rolenum': datasets.Value(
|
| 116 |
-
'role': datasets.Value(
|
| 117 |
-
'initiated_dialogue': datasets.Value(
|
| 118 |
-
'turn_number': datasets.Value(
|
| 119 |
-
'lang': datasets.Value(
|
| 120 |
}
|
| 121 |
)
|
| 122 |
}
|
|
@@ -125,11 +133,11 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 125 |
[
|
| 126 |
datasets.features.Sequence(
|
| 127 |
{
|
| 128 |
-
'id': datasets.Value(
|
| 129 |
-
'orig': datasets.Value(
|
| 130 |
-
'norm': datasets.Value(
|
| 131 |
-
'mt': datasets.Value(
|
| 132 |
-
'ref': datasets.Value(
|
| 133 |
'utterance_meta': datasets.features.Sequence(
|
| 134 |
{
|
| 135 |
'judgment': ClassLabel(num_classes=3, names=['poor', 'medium', 'perfect']),
|
|
@@ -182,6 +190,12 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 182 |
dialogue_info = {k: dialogue[k] for k in dialogue_info_keys}
|
| 183 |
if dialogue_info['end_time'] is None:
|
| 184 |
dialogue_info['end_time'] = ''
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
# Main data: the utterances
|
| 186 |
for utterance_id in dialogue['utterances']:
|
| 187 |
utterance = dialogue['utterances'][utterance_id]
|
|
@@ -205,15 +219,7 @@ class Diabla(datasets.GeneratorBasedBuilder):
|
|
| 205 |
'ref': reference_text,
|
| 206 |
'utterance_meta': utterance_info
|
| 207 |
}
|
| 208 |
-
|
| 209 |
-
"verbatim_quality": datasets.Value("string"),
|
| 210 |
-
|
| 211 |
-
"particular_problems": "",
|
| 212 |
-
"tech": "There were no technical problems",
|
| 213 |
-
"would_use": false,
|
| 214 |
-
"timestamp": "2018-05-18T17:11:39.104833",
|
| 215 |
-
|
| 216 |
-
"technical_issue": "There were no technical problems"
|
| 217 |
# add to history (without dialogue info and history)
|
| 218 |
dialogue_history.append(utterance_instance.copy())
|
| 219 |
utterance_instance['dialogue_meta'] = dialogue_info
|
|
|
|
| 50 |
|
| 51 |
BUILDER_CONFIGS = [
|
| 52 |
DiablaConfig(
|
| 53 |
+
name='plain_text',
|
| 54 |
+
version=datasets.Version('1.0.0", ''),
|
| 55 |
+
description='Plain text',
|
| 56 |
),
|
| 57 |
]
|
| 58 |
|
|
|
|
| 62 |
description=_DESCRIPTION,
|
| 63 |
features=datasets.Features(
|
| 64 |
{
|
| 65 |
+
'id': datasets.Value('string'),
|
| 66 |
+
'orig': datasets.Value('string'),
|
| 67 |
+
'norm': datasets.Value('string'),
|
| 68 |
+
'mt': datasets.Value('string'),
|
| 69 |
+
'ref': datasets.Value('string'),
|
| 70 |
'utterance_meta': datasets.features.Sequence(
|
| 71 |
{
|
| 72 |
'eval-judgment': ClassLabel(num_classes=3, names=['poor', 'medium', 'perfect']),
|
|
|
|
| 81 |
),
|
| 82 |
'dialogue_meta': datasets.features.Sequence(
|
| 83 |
{
|
| 84 |
+
'start_time': datasets.Value('string'),
|
| 85 |
+
'end_time' : datasets.Value('string'),
|
| 86 |
+
'translation_model': datasets.Value('string'),
|
| 87 |
'final_evaluation_user1': datasets.features.Sequence(
|
| 88 |
{
|
| 89 |
+
'style': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 90 |
+
'coherence': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 91 |
+
'grammaticality': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 92 |
+
'meaning': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 93 |
+
'word_choice': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent'])
|
| 94 |
+
}
|
| 95 |
+
),
|
| 96 |
+
'final_evaluation_user2': datasets.features.Sequence(
|
| 97 |
+
{
|
| 98 |
+
'style': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 99 |
+
'coherence': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 100 |
+
'grammaticality': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 101 |
+
'meaning': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent']),
|
| 102 |
+
'word_choice': ClassLabel(num_classes=3, names=['poor', 'average', 'good', 'excellent'])
|
| 103 |
+
}
|
| 104 |
+
),
|
| 105 |
'scenario': datasets.features.Sequence(
|
| 106 |
[
|
| 107 |
[
|
|
|
|
| 111 |
),
|
| 112 |
'user1': datasets.features.Sequence(
|
| 113 |
{
|
| 114 |
+
'rolenum': datasets.Value('int64'),
|
| 115 |
+
'role': datasets.Value('string'),
|
| 116 |
+
'initiated_dialogue': datasets.Value('bool'),
|
| 117 |
+
'turn_number': datasets.Value('int64'),
|
| 118 |
+
'lang': datasets.Value('string'),
|
| 119 |
}
|
| 120 |
),
|
| 121 |
'user2': datasets.features.Sequence(
|
| 122 |
{
|
| 123 |
+
'rolenum': datasets.Value('int64'),
|
| 124 |
+
'role': datasets.Value('string'),
|
| 125 |
+
'initiated_dialogue': datasets.Value('bool'),
|
| 126 |
+
'turn_number': datasets.Value('int64'),
|
| 127 |
+
'lang': datasets.Value('string'),
|
| 128 |
}
|
| 129 |
)
|
| 130 |
}
|
|
|
|
| 133 |
[
|
| 134 |
datasets.features.Sequence(
|
| 135 |
{
|
| 136 |
+
'id': datasets.Value('string'),
|
| 137 |
+
'orig': datasets.Value('string'),
|
| 138 |
+
'norm': datasets.Value('string'),
|
| 139 |
+
'mt': datasets.Value('string'),
|
| 140 |
+
'ref': datasets.Value('string'),
|
| 141 |
'utterance_meta': datasets.features.Sequence(
|
| 142 |
{
|
| 143 |
'judgment': ClassLabel(num_classes=3, names=['poor', 'medium', 'perfect']),
|
|
|
|
| 190 |
dialogue_info = {k: dialogue[k] for k in dialogue_info_keys}
|
| 191 |
if dialogue_info['end_time'] is None:
|
| 192 |
dialogue_info['end_time'] = ''
|
| 193 |
+
for info_to_remove in ['interface','verbatim_quality',
|
| 194 |
+
'particular_problems', 'tech',
|
| 195 |
+
'would_use', 'timestamp', 'technical_issue']:
|
| 196 |
+
del dialogue_info['final_evaluation_user1'][info_to_remove]
|
| 197 |
+
del dialogue_info['final_evaluation_user2'][info_to_remove]
|
| 198 |
+
|
| 199 |
# Main data: the utterances
|
| 200 |
for utterance_id in dialogue['utterances']:
|
| 201 |
utterance = dialogue['utterances'][utterance_id]
|
|
|
|
| 219 |
'ref': reference_text,
|
| 220 |
'utterance_meta': utterance_info
|
| 221 |
}
|
| 222 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 223 |
# add to history (without dialogue info and history)
|
| 224 |
dialogue_history.append(utterance_instance.copy())
|
| 225 |
utterance_instance['dialogue_meta'] = dialogue_info
|