ohashi56225
commited on
Commit
•
b7050fa
1
Parent(s):
5aeb3eb
Update README.md
Browse files
README.md
CHANGED
@@ -72,15 +72,20 @@ dataset_info:
|
|
72 |
- name: value
|
73 |
dtype: string
|
74 |
- name: db_result
|
75 |
-
|
76 |
- name: candidate_entities
|
77 |
-
sequence:
|
|
|
|
|
|
|
78 |
- name: active_entity
|
79 |
sequence:
|
80 |
- name: slot
|
81 |
dtype: string
|
|
|
82 |
- name: value
|
83 |
dtype: string
|
|
|
84 |
- name: book_result
|
85 |
sequence:
|
86 |
- name: domain
|
@@ -91,14 +96,145 @@ dataset_info:
|
|
91 |
dtype: string
|
92 |
splits:
|
93 |
- name: train
|
94 |
-
num_bytes:
|
95 |
num_examples: 3646
|
96 |
- name: validation
|
97 |
-
num_bytes:
|
98 |
num_examples: 300
|
99 |
- name: test
|
100 |
-
num_bytes:
|
101 |
num_examples: 300
|
102 |
download_size: 11016438
|
103 |
-
dataset_size:
|
104 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
- name: value
|
73 |
dtype: string
|
74 |
- name: db_result
|
75 |
+
struct:
|
76 |
- name: candidate_entities
|
77 |
+
sequence:
|
78 |
+
dtype: string
|
79 |
+
id: entity_name
|
80 |
+
id: candidate_entities
|
81 |
- name: active_entity
|
82 |
sequence:
|
83 |
- name: slot
|
84 |
dtype: string
|
85 |
+
id: active_entity/slot
|
86 |
- name: value
|
87 |
dtype: string
|
88 |
+
id: active_entity/value
|
89 |
- name: book_result
|
90 |
sequence:
|
91 |
- name: domain
|
|
|
96 |
dtype: string
|
97 |
splits:
|
98 |
- name: train
|
99 |
+
num_bytes: 60731411
|
100 |
num_examples: 3646
|
101 |
- name: validation
|
102 |
+
num_bytes: 5000420
|
103 |
num_examples: 300
|
104 |
- name: test
|
105 |
+
num_bytes: 5085276
|
106 |
num_examples: 300
|
107 |
download_size: 11016438
|
108 |
+
dataset_size: 70817107
|
109 |
---
|
110 |
+
|
111 |
+
# Dataset Card for JMultiWOZ
|
112 |
+
|
113 |
+
## Dataset Description
|
114 |
+
|
115 |
+
- **Repository:** [nu-dialouge/jmultiwoz](https://github.com/nu-dialogue/jmultiwoz)
|
116 |
+
- **Paper:** [JMultiWOZ: A Large-Scale Japanese Multi-Domain Task-Oriented Dialogue Dataset]()
|
117 |
+
- **Point of Contact:** [Atsumoto Ohashi](ohashi.atsumoto.c0@s.mail.nagoya-u.ac.jp)
|
118 |
+
|
119 |
+
### Dataset Summary
|
120 |
+
|
121 |
+
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.
|
122 |
+
|
123 |
+
### Supported Tasks
|
124 |
+
|
125 |
+
- **Dialogue State Tracking**: The dataset can be used to train models for dialogue state tracking, which is the task of predicting the user's belief state at each turn in the dialogue.
|
126 |
+
- **Dialogue Generation**: The dataset can be used to train models for dialogue generation, which is the task of generating a response given the dialogue history.
|
127 |
+
|
128 |
+
### Languages
|
129 |
+
|
130 |
+
The text in the dataset is in Japanese (`ja`).
|
131 |
+
|
132 |
+
## Dataset Usage
|
133 |
+
```python
|
134 |
+
from datasets import load_dataset
|
135 |
+
|
136 |
+
dataset = load_dataset("nu-dialogue/jmultiwoz", trust_remote_code=True)
|
137 |
+
```
|
138 |
+
|
139 |
+
## Dataset Structure
|
140 |
+
|
141 |
+
### Data Instances
|
142 |
+
|
143 |
+
A data instance is a full multi-turn dialogue between a `USER` and a `SYSTEM`. Each turn has an `utterance`:
|
144 |
+
```json
|
145 |
+
[
|
146 |
+
"福岡へ行くよていなのですが、値段が普通くらいの宿泊施設を探してもらっていいですか?",
|
147 |
+
"かしこまりました。ではWITH THE STYLE FUKUOKAはいかがでしょうか。"
|
148 |
+
]
|
149 |
+
```
|
150 |
+
|
151 |
+
`SYSTEM` turn also has a `dialogue_state` which contains `belief_state`, `book_state`, `db_result`, and `book_result`:
|
152 |
+
|
153 |
+
`belief_state`:
|
154 |
+
```json
|
155 |
+
{
|
156 |
+
"domain": ["general", "general", "hotel", ...],
|
157 |
+
"slot": ["active_domain", "city", "pricerange", ...],
|
158 |
+
"value": ["hotel", "福岡", "普通", ...]
|
159 |
+
}
|
160 |
+
```
|
161 |
+
|
162 |
+
`book_state`:
|
163 |
+
```json
|
164 |
+
{
|
165 |
+
"domain": ["hotel", "hotel", "hotel", ...],
|
166 |
+
"slot": ["people", "day", "stay", ...],
|
167 |
+
"value": [None, None, None, ...]
|
168 |
+
}
|
169 |
+
```
|
170 |
+
|
171 |
+
`db_result`:
|
172 |
+
```json
|
173 |
+
{
|
174 |
+
"candidate_entities": ["WITH THE STYLE FUKUOKA", "ANA クラウンプラザホテル福岡", ...],
|
175 |
+
"active_entity": {
|
176 |
+
"slot": ["city", "name", "genre", ...],
|
177 |
+
"value": ["福岡", "WITH THE STYLE FUKUOKA", "リゾートホテル", ...]
|
178 |
+
}
|
179 |
+
```
|
180 |
+
|
181 |
+
### Data Fields
|
182 |
+
|
183 |
+
Each dialogue instance has the following fields:
|
184 |
+
- `dialogue_id` (int32): A unique identifier for the dialogue.
|
185 |
+
- `dialogue_name` (string): A name for the dialogue.
|
186 |
+
- `system_name` (string): The name of the wizard.
|
187 |
+
- `user_name` (string): The name of the user.
|
188 |
+
- `goal` (sequence): The user's goal for the dialogue.
|
189 |
+
- `domain` (string): The domain of the goal.
|
190 |
+
- `task` (string): The task of the goal.
|
191 |
+
- `slot` (string): The slot of the goal.
|
192 |
+
- `value` (string): The value of the goal.
|
193 |
+
- `goal_description` (sequence): A description of the user's goal.
|
194 |
+
- `domain` (string): The domain of the goal.
|
195 |
+
- `text` (string): The text of the goal.
|
196 |
+
- `turns` (sequence): The turns in the dialogue.
|
197 |
+
- `turn_id` (int32): A unique identifier for the turn.
|
198 |
+
- `speaker` (string): The speaker of the turn.
|
199 |
+
- `utterance` (string): The utterance of the turn.
|
200 |
+
- `dialogue_state` (struct): The dialogue state of the turn.
|
201 |
+
- `belief_state` (sequence): The belief state of the turn.
|
202 |
+
- `domain` (string): The domain of the belief state.
|
203 |
+
- `slot` (string): The slot of the belief state.
|
204 |
+
- `value` (string): The value of the belief state.
|
205 |
+
- `book_state` (sequence): The book state of the turn.
|
206 |
+
- `domain` (string): The domain of the book state.
|
207 |
+
- `slot` (string): The slot of the book state.
|
208 |
+
- `value` (string): The value of the book state.
|
209 |
+
- `db_result` (struct): The database result of the turn.
|
210 |
+
- `candidate_entities` (sequence): The candidate entities of the database result.
|
211 |
+
- `entity_name` (string): The name of the entity.
|
212 |
+
- `active_entity` (sequence): The active entity of the database result.
|
213 |
+
- `slot` (string): The slot of the active entity.
|
214 |
+
- `value` (string): The value of the active entity.
|
215 |
+
- `book_result` (sequence): The book result of the turn.
|
216 |
+
- `domain` (string): The domain of the book result.
|
217 |
+
- `success` (string): The success of the book result.
|
218 |
+
- `ref` (string): The reference of the book result.
|
219 |
+
|
220 |
+
### Data Splits
|
221 |
+
|
222 |
+
The dataset is split into a train, validation, and test split with the following sizes:
|
223 |
+
|
224 |
+
| | train | validation | test |
|
225 |
+
|--- | ---: | ---: | ---: |
|
226 |
+
| Number of dialogues | 3646 | 300 | 300 |
|
227 |
+
| Number of turns | 52,405 | 4,346 | 4,435 |
|
228 |
+
|
229 |
+
|
230 |
+
## Citation Information
|
231 |
+
```bibtex
|
232 |
+
@inproceedings{ohashi-etal-2024-jmultiwoz,
|
233 |
+
title = "JMultiWOZ: A Large-Scale Japanese Multi-Domain Task-Oriented Dialogue Dataset",
|
234 |
+
author = "Ohashi, Atsumoto and Hirai, Ryu and Iizuka, Shinya and Higashinaka, Ryuichiro",
|
235 |
+
booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation",
|
236 |
+
year = "2024",
|
237 |
+
url = "",
|
238 |
+
pages = "",
|
239 |
+
}
|
240 |
+
```
|