Datasets:
ShutongFeng
commited on
Commit
•
de988d8
1
Parent(s):
5638297
Update emowoz.py
Browse filesCleaning up comments.
emowoz.py
CHANGED
@@ -11,19 +11,13 @@
|
|
11 |
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
# See the License for the specific language governing permissions and
|
13 |
# limitations under the License.
|
14 |
-
# TODO: Address all TODOs and remove all explanatory comments
|
15 |
-
"""TODO: Add a description here."""
|
16 |
|
17 |
|
18 |
-
import csv
|
19 |
import json
|
20 |
-
import os
|
21 |
|
22 |
import datasets
|
23 |
|
24 |
|
25 |
-
# TODO: Add BibTeX citation
|
26 |
-
# Find for instance the citation on arxiv or on the dataset repo/website
|
27 |
_CITATION = """\
|
28 |
@inproceedings{feng-etal-2022-emowoz,
|
29 |
title = "{E}mo{WOZ}: A Large-Scale Corpus and Labelling Scheme for Emotion Recognition in Task-Oriented Dialogue Systems",
|
@@ -41,12 +35,23 @@ _CITATION = """\
|
|
41 |
publisher = "European Language Resources Association",
|
42 |
url = "https://aclanthology.org/2022.lrec-1.436",
|
43 |
pages = "4096--4113",
|
44 |
-
abstract = "The ability to recognise emotions lends a conversational artificial intelligence a human
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
}
|
46 |
"""
|
47 |
|
48 |
-
# TODO: Add description of the dataset here
|
49 |
-
# You can copy an official description
|
50 |
_DESCRIPTION = """\
|
51 |
EmoWOZ is a user emotion recognition in task-oriented dialogues dataset, \
|
52 |
consisting all dialogues from MultiWOZ and 1000 additional human-machine \
|
@@ -57,15 +62,10 @@ following emotions: 0: neutral, 1: fearful, 2: dissatisfied, 3: apologetic, \
|
|
57 |
dataset homepage.
|
58 |
"""
|
59 |
|
60 |
-
# TODO: Add a link to an official homepage for the dataset here
|
61 |
_HOMEPAGE = "https://zenodo.org/record/6506504"
|
62 |
|
63 |
-
# TODO: Add the licence for the dataset here if you can find it
|
64 |
_LICENSE = "https://creativecommons.org/licenses/by-nc/4.0/legalcode"
|
65 |
|
66 |
-
# TODO: Add link to the official dataset URLs here
|
67 |
-
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
68 |
-
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
69 |
_URLS = {
|
70 |
"emowoz_multiwoz": "https://zenodo.org/record/6506504/files/emowoz-multiwoz.json",
|
71 |
"emowoz_dialmage": "https://zenodo.org/record/6506504/files/emowoz-dialmage.json",
|
@@ -73,34 +73,22 @@ _URLS = {
|
|
73 |
}
|
74 |
|
75 |
|
76 |
-
|
77 |
class EmoWOZ(datasets.GeneratorBasedBuilder):
|
78 |
"""EmoWOZ: A Large-Scale Corpus and Labelling Scheme for Emotion Recognition in Task-Oriented Dialogue Systems"""
|
79 |
|
80 |
VERSION = datasets.Version("1.0.0")
|
81 |
|
82 |
-
# This is an example of a dataset with multiple configurations.
|
83 |
-
# If you don't want/need to define several sub-sets in your dataset,
|
84 |
-
# just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
|
85 |
-
|
86 |
-
# If you need to make complex sub-parts in the datasets with configurable options
|
87 |
-
# You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
|
88 |
-
# BUILDER_CONFIG_CLASS = MyBuilderConfig
|
89 |
-
|
90 |
-
# You will be able to load one or the other configurations in the following list with
|
91 |
-
# data = datasets.load_dataset('my_dataset', 'first_domain')
|
92 |
-
# data = datasets.load_dataset('my_dataset', 'second_domain')
|
93 |
BUILDER_CONFIGS = [
|
94 |
datasets.BuilderConfig(name="emowoz", version=VERSION, description="This part contains all user-emotion-annotated dialogues from EmoWOZ"),
|
95 |
datasets.BuilderConfig(name="multiwoz", version=VERSION, description="This part contains all user-emotion-annotated dialogues from MultiWOZ"),
|
96 |
datasets.BuilderConfig(name="dialmage", version=VERSION, description="This part contains all user-emotion-annotated dialogues from human-machine interactions (DialMAGE)"),
|
97 |
]
|
98 |
|
99 |
-
|
100 |
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
features = datasets.Features(
|
105 |
{
|
106 |
"dialogue_id": datasets.Value("string"),
|
@@ -121,19 +109,14 @@ class EmoWOZ(datasets.GeneratorBasedBuilder):
|
|
121 |
citation=_CITATION,
|
122 |
)
|
123 |
|
|
|
124 |
def _split_generators(self, dl_manager):
|
125 |
-
|
126 |
-
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
127 |
-
|
128 |
-
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
|
129 |
-
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
130 |
-
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
131 |
data_dir = dl_manager.download_and_extract(_URLS)
|
132 |
|
133 |
return [
|
134 |
datasets.SplitGenerator(
|
135 |
name=datasets.Split.TRAIN,
|
136 |
-
# These kwargs will be passed to _generate_examples
|
137 |
gen_kwargs={
|
138 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
139 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
@@ -143,7 +126,6 @@ class EmoWOZ(datasets.GeneratorBasedBuilder):
|
|
143 |
),
|
144 |
datasets.SplitGenerator(
|
145 |
name=datasets.Split.VALIDATION,
|
146 |
-
# These kwargs will be passed to _generate_examples
|
147 |
gen_kwargs={
|
148 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
149 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
@@ -153,7 +135,6 @@ class EmoWOZ(datasets.GeneratorBasedBuilder):
|
|
153 |
),
|
154 |
datasets.SplitGenerator(
|
155 |
name=datasets.Split.TEST,
|
156 |
-
# These kwargs will be passed to _generate_examples
|
157 |
gen_kwargs={
|
158 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
159 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
@@ -163,10 +144,9 @@ class EmoWOZ(datasets.GeneratorBasedBuilder):
|
|
163 |
)
|
164 |
]
|
165 |
|
166 |
-
|
167 |
def _generate_examples(self, multiwoz_filepath, dialmage_filepath, split_filepath, split):
|
168 |
-
|
169 |
-
# The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
|
170 |
with open(multiwoz_filepath, encoding="utf-8") as f:
|
171 |
multiwoz_dialogues = json.load(f)
|
172 |
with open(dialmage_filepath, encoding="utf-8") as f:
|
|
|
11 |
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
# See the License for the specific language governing permissions and
|
13 |
# limitations under the License.
|
|
|
|
|
14 |
|
15 |
|
|
|
16 |
import json
|
|
|
17 |
|
18 |
import datasets
|
19 |
|
20 |
|
|
|
|
|
21 |
_CITATION = """\
|
22 |
@inproceedings{feng-etal-2022-emowoz,
|
23 |
title = "{E}mo{WOZ}: A Large-Scale Corpus and Labelling Scheme for Emotion Recognition in Task-Oriented Dialogue Systems",
|
|
|
35 |
publisher = "European Language Resources Association",
|
36 |
url = "https://aclanthology.org/2022.lrec-1.436",
|
37 |
pages = "4096--4113",
|
38 |
+
abstract = "The ability to recognise emotions lends a conversational artificial intelligence a human \
|
39 |
+
touch. While emotions in chit-chat dialogues have received substantial attention, emotions in \
|
40 |
+
task-oriented dialogues remain largely unaddressed. This is despite emotions and dialogue success \
|
41 |
+
having equally important roles in a natural system. Existing emotion-annotated task-oriented corpora \
|
42 |
+
are limited in size, label richness, and public availability, creating a bottleneck for downstream \
|
43 |
+
tasks. To lay a foundation for studies on emotions in task-oriented dialogues, we introduce EmoWOZ, a \
|
44 |
+
large-scale manually emotion-annotated corpus of task-oriented dialogues. EmoWOZ is based on MultiWOZ, \
|
45 |
+
a multi-domain task-oriented dialogue dataset. It contains more than 11K dialogues with more than 83K \
|
46 |
+
emotion annotations of user utterances. In addition to Wizard-of-Oz dialogues from MultiWOZ, we collect \
|
47 |
+
human-machine dialogues within the same set of domains to sufficiently cover the space of various emotions \
|
48 |
+
that can happen during the lifetime of a data-driven dialogue system. To the best of our knowledge, this \
|
49 |
+
is the first large-scale open-source corpus of its kind. We propose a novel emotion labelling scheme, \
|
50 |
+
which is tailored to task-oriented dialogues. We report a set of experimental results to show the usability \
|
51 |
+
of this corpus for emotion recognition and state tracking in task-oriented dialogues.",
|
52 |
}
|
53 |
"""
|
54 |
|
|
|
|
|
55 |
_DESCRIPTION = """\
|
56 |
EmoWOZ is a user emotion recognition in task-oriented dialogues dataset, \
|
57 |
consisting all dialogues from MultiWOZ and 1000 additional human-machine \
|
|
|
62 |
dataset homepage.
|
63 |
"""
|
64 |
|
|
|
65 |
_HOMEPAGE = "https://zenodo.org/record/6506504"
|
66 |
|
|
|
67 |
_LICENSE = "https://creativecommons.org/licenses/by-nc/4.0/legalcode"
|
68 |
|
|
|
|
|
|
|
69 |
_URLS = {
|
70 |
"emowoz_multiwoz": "https://zenodo.org/record/6506504/files/emowoz-multiwoz.json",
|
71 |
"emowoz_dialmage": "https://zenodo.org/record/6506504/files/emowoz-dialmage.json",
|
|
|
73 |
}
|
74 |
|
75 |
|
|
|
76 |
class EmoWOZ(datasets.GeneratorBasedBuilder):
|
77 |
"""EmoWOZ: A Large-Scale Corpus and Labelling Scheme for Emotion Recognition in Task-Oriented Dialogue Systems"""
|
78 |
|
79 |
VERSION = datasets.Version("1.0.0")
|
80 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
BUILDER_CONFIGS = [
|
82 |
datasets.BuilderConfig(name="emowoz", version=VERSION, description="This part contains all user-emotion-annotated dialogues from EmoWOZ"),
|
83 |
datasets.BuilderConfig(name="multiwoz", version=VERSION, description="This part contains all user-emotion-annotated dialogues from MultiWOZ"),
|
84 |
datasets.BuilderConfig(name="dialmage", version=VERSION, description="This part contains all user-emotion-annotated dialogues from human-machine interactions (DialMAGE)"),
|
85 |
]
|
86 |
|
87 |
+
DEFAULT_CONFIG_NAME = "emowoz"
|
88 |
|
89 |
+
|
90 |
+
def _info(self):
|
91 |
+
|
92 |
features = datasets.Features(
|
93 |
{
|
94 |
"dialogue_id": datasets.Value("string"),
|
|
|
109 |
citation=_CITATION,
|
110 |
)
|
111 |
|
112 |
+
|
113 |
def _split_generators(self, dl_manager):
|
114 |
+
|
|
|
|
|
|
|
|
|
|
|
115 |
data_dir = dl_manager.download_and_extract(_URLS)
|
116 |
|
117 |
return [
|
118 |
datasets.SplitGenerator(
|
119 |
name=datasets.Split.TRAIN,
|
|
|
120 |
gen_kwargs={
|
121 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
122 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
|
|
126 |
),
|
127 |
datasets.SplitGenerator(
|
128 |
name=datasets.Split.VALIDATION,
|
|
|
129 |
gen_kwargs={
|
130 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
131 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
|
|
135 |
),
|
136 |
datasets.SplitGenerator(
|
137 |
name=datasets.Split.TEST,
|
|
|
138 |
gen_kwargs={
|
139 |
"multiwoz_filepath": data_dir['emowoz_multiwoz'],
|
140 |
"dialmage_filepath": data_dir['emowoz_dialmage'],
|
|
|
144 |
)
|
145 |
]
|
146 |
|
147 |
+
|
148 |
def _generate_examples(self, multiwoz_filepath, dialmage_filepath, split_filepath, split):
|
149 |
+
|
|
|
150 |
with open(multiwoz_filepath, encoding="utf-8") as f:
|
151 |
multiwoz_dialogues = json.load(f)
|
152 |
with open(dialmage_filepath, encoding="utf-8") as f:
|