Datasets:

jmultiwoz / jmultiwoz.py
ohashi56225's picture
Update URL
b41a143 verified
raw
history blame
12.2 kB
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# TODO: Address all TODOs and remove all explanatory comments
"""JMultiWOZ: Japanese Multi-Domain Wizard-of-Oz dataset for task-oriented dialogue modelling"""
import json
import os
import datasets
# TODO: Add BibTeX citation
# Find for instance the citation on arxiv or on the dataset repo/website
_CITATION = """\
@inproceedings{ohashi-etal-2024-jmultiwoz,
title = "JMultiWOZ: A Large-Scale Japanese Multi-Domain Task-Oriented Dialogue Dataset",
author = "Ohashi, Atsumoto and Hirai, Ryu and Iizuka, Shinya and Higashinaka, Ryuichiro",
booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation",
year = "2024",
url = "",
pages = "",
}
"""
# TODO: Add description of the dataset here
# You can copy an official description
_DESCRIPTION = """\
JMultiWOZ is a large-scale Japanese multi-domain task-oriented dialogue dataset. The dataset is collected using
the Wizard-of-Oz (WoZ) methodology, where two human annotators simulate the user and the system. The dataset contains
4,246 dialogues across 6 domains, including restaurant, hotel, attraction, shopping, taxi, and weather. Available
annotations include user goal, dialogue state, and utterances.
"""
# TODO: Add a link to an official homepage for the dataset here
_HOMEPAGE = "https://github.com/nu-dialogue/jmultiwoz"
# TODO: Add the licence for the dataset here if you can find it
_LICENSE = "CC BY-ND 4.0"
# TODO: Add link to the official dataset URLs here
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
_URLS = {
"original_zip": "https://github.com/nu-dialogue/jmultiwoz/raw/master/dataset/JMultiWOZ_1.0.zip",
}
def _flatten_value(values) -> str:
if not isinstance(values, list):
return values
flat_values = [
_flatten_value(v) if isinstance(v, list) else v for v in values
]
return "[" + ", ".join(flat_values) + "]"
# TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
class JMultiWOZDataset(datasets.GeneratorBasedBuilder):
VERSION = datasets.Version("1.0.0")
def _info(self):
# TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
features = datasets.Features({
"dialogue_id": datasets.Value("int32"),
"dialogue_name": datasets.Value("string"),
"system_name": datasets.Value("string"),
"user_name": datasets.Value("string"),
"goal": datasets.Sequence({
"domain": datasets.Value("string"),
"task": datasets.Value("string"),
"slot": datasets.Value("string"),
"value": datasets.Value("string"),
}),
"goal_description": datasets.Sequence({
"domain": datasets.Value("string"),
"text": datasets.Value("string"),
}),
"turns": datasets.Sequence({
"turn_id": datasets.Value("int32"),
"speaker": datasets.Value("string"),
"utterance": datasets.Value("string"),
"dialogue_state": {
"belief_state": datasets.Sequence({
"domain": datasets.Value("string"),
"slot": datasets.Value("string"),
"value": datasets.Value("string"),
}),
"book_state": datasets.Sequence({
"domain": datasets.Value("string"),
"slot": datasets.Value("string"),
"value": datasets.Value("string"),
}),
"db_result": {
"candidate_entities": datasets.Sequence(datasets.Value("string")),
"active_entity": datasets.Sequence({
"slot": datasets.Value("string"),
"value": datasets.Value("string"),
})
},
"book_result": datasets.Sequence({
"domain": datasets.Value("string"),
"success": datasets.Value("string"),
"ref": datasets.Value("string"),
}),
}
}),
})
return datasets.DatasetInfo(
# This is the description that will appear on the datasets page.
description=_DESCRIPTION,
# This defines the different columns of the dataset and their types
features=features, # Here we define them above because they are different between the two configurations
# If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
# specify them. They'll be used if as_supervised=True in builder.as_dataset.
# supervised_keys=("sentence", "label"),
# Homepage of the dataset for documentation
homepage=_HOMEPAGE,
# License for the dataset if available
license=_LICENSE,
# Citation for the dataset
citation=_CITATION,
)
def _split_generators(self, dl_manager):
# TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
# 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.
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
data_dir = dl_manager.download_and_extract(_URLS["original_zip"])
split_list_path = os.path.join(data_dir, "JMultiWOZ_1.0/split_list.json")
dialogues_path = os.path.join(data_dir, "JMultiWOZ_1.0/dialogues.json")
with open(split_list_path, "r", encoding="utf-8") as f:
split_list = json.load(f)
with open(dialogues_path, "r", encoding="utf-8") as f:
dialogues = json.load(f)
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
# These kwargs will be passed to _generate_examples
gen_kwargs={
"dialogues": [dialogues[dialogue_name] for dialogue_name in split_list["train"]],
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
# These kwargs will be passed to _generate_examples
gen_kwargs={
"dialogues": [dialogues[dialogue_name] for dialogue_name in split_list["dev"]],
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
# These kwargs will be passed to _generate_examples
gen_kwargs={
"dialogues": [dialogues[dialogue_name] for dialogue_name in split_list["test"]],
},
),
]
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
def _generate_examples(self, dialogues):
# TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
# The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.
for id_, dialogue in enumerate(dialogues):
example = {
"dialogue_id": dialogue["dialogue_id"],
"dialogue_name": dialogue["dialogue_name"],
"system_name": dialogue["system_name"],
"user_name": dialogue["user_name"],
"goal": [],
"goal_description": [],
"turns": [],
}
for domain, tasks in dialogue["goal"].items():
for task, slot_values in tasks.items():
if task == "reqt":
slot_values = {slot: None for slot in slot_values}
for slot, value in slot_values.items():
example["goal"].append({
"domain": domain,
"task": task,
"slot": slot,
"value": value,
})
for domain, texts in dialogue["goal_description"].items():
for text in texts:
example["goal_description"].append({
"domain": domain,
"text": text,
})
for turn in dialogue["turns"]:
example_turn = {
"turn_id": turn["turn_id"],
"speaker": turn["speaker"],
"utterance": turn["utterance"],
"dialogue_state": {
"belief_state": [],
"book_state": [],
"db_result": {},
"book_result": [],
},
}
if turn["speaker"] == "SYSTEM":
for domain, slots in turn["dialogue_state"]["belief_state"].items():
for slot, value in slots.items():
example_turn["dialogue_state"]["belief_state"].append({
"domain": domain,
"slot": slot,
"value": value,
})
for domain, slots in turn["dialogue_state"]["book_state"].items():
for slot, value in slots.items():
example_turn["dialogue_state"]["book_state"].append({
"domain": domain,
"slot": slot,
"value": value,
})
candidate_entities = turn["dialogue_state"]["db_result"]["candidate_entities"]
active_entity = turn["dialogue_state"]["db_result"]["active_entity"]
if not active_entity:
active_entity = {}
example_turn["dialogue_state"]["db_result"] = {
"candidate_entities":candidate_entities,
"active_entity": [{
"slot": slot,
"value": _flatten_value(value),
} for slot, value in active_entity.items()]
}
for domain, result in turn["dialogue_state"]["book_result"].items():
example_turn["dialogue_state"]["book_result"].append({
"domain": domain,
"success": result["success"],
"ref": result["ref"],
})
example["turns"].append(example_turn)
yield id_, example