Datasets:
Commit
·
026fd78
0
Parent(s):
Update files from the datasets library (from 1.2.0)
Browse filesRelease notes: https://github.com/huggingface/datasets/releases/tag/1.2.0
- .gitattributes +27 -0
- README.md +225 -0
- dataset_infos.json +1 -0
- dummy/dialogues/0.0.0/dummy_data.zip +3 -0
- dummy/tasks/0.0.0/dummy_data.zip +3 -0
- meta_woz.py +154 -0
.gitattributes
ADDED
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.bin.* filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
18 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
19 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
20 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
21 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
22 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
23 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
24 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
25 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
26 |
+
*.zstandard filter=lfs diff=lfs merge=lfs -text
|
27 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
@@ -0,0 +1,225 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
annotations_creators:
|
3 |
+
- crowdsourced
|
4 |
+
language_creators:
|
5 |
+
- crowdsourced
|
6 |
+
languages:
|
7 |
+
- en
|
8 |
+
licenses:
|
9 |
+
- Microsoft Research Data License Agreement
|
10 |
+
multilinguality:
|
11 |
+
- monolingual
|
12 |
+
size_categories:
|
13 |
+
- 10K<n<100K
|
14 |
+
source_datasets:
|
15 |
+
- original
|
16 |
+
task_categories:
|
17 |
+
- sequence-modeling
|
18 |
+
task_ids:
|
19 |
+
- dialogue-modeling
|
20 |
+
---
|
21 |
+
|
22 |
+
# Dataset Card for MetaLWOz
|
23 |
+
|
24 |
+
## Table of Contents
|
25 |
+
- [Dataset Description](#dataset-description)
|
26 |
+
- [Dataset Summary](#dataset-summary)
|
27 |
+
- [Supported Tasks](#supported-tasks-and-leaderboards)
|
28 |
+
- [Languages](#languages)
|
29 |
+
- [Dataset Structure](#dataset-structure)
|
30 |
+
- [Data Instances](#data-instances)
|
31 |
+
- [Data Fields](#data-instances)
|
32 |
+
- [Data Splits](#data-instances)
|
33 |
+
- [Dataset Creation](#dataset-creation)
|
34 |
+
- [Curation Rationale](#curation-rationale)
|
35 |
+
- [Source Data](#source-data)
|
36 |
+
- [Annotations](#annotations)
|
37 |
+
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
38 |
+
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
39 |
+
- [Social Impact of Dataset](#social-impact-of-dataset)
|
40 |
+
- [Discussion of Biases](#discussion-of-biases)
|
41 |
+
- [Other Known Limitations](#other-known-limitations)
|
42 |
+
- [Additional Information](#additional-information)
|
43 |
+
- [Dataset Curators](#dataset-curators)
|
44 |
+
- [Licensing Information](#licensing-information)
|
45 |
+
- [Citation Information](#citation-information)
|
46 |
+
|
47 |
+
## Dataset Description
|
48 |
+
|
49 |
+
- **Repository:** [MetaLWOz Project Website](https://www.microsoft.com/en-us/research/project/metalwoz/)
|
50 |
+
- **Paper:** [Fast Domain Adaptation for Goal-Oriented Dialogue Using a Hybrid Generative-Retrieval Transformer](https://ieeexplore.ieee.org/abstract/document/9053599), and [Hybrid Generative-Retrieval Transformers for Dialogue Domain Adaptation](https://arxiv.org/pdf/2003.01680.pdf)
|
51 |
+
- **Point of Contact:** [Hannes Schulz](https://www.microsoft.com/en-us/research/people/haschulz/)
|
52 |
+
|
53 |
+
### Dataset Summary
|
54 |
+
|
55 |
+
MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models.
|
56 |
+
We introduce the Meta-Learning Wizard of Oz (MetaLWOz) dialogue dataset for developing fast adaptation methods for
|
57 |
+
conversation models. This data can be used to train task-oriented dialogue models, specifically to develop methods to
|
58 |
+
quickly simulate user responses with a small amount of data. Such fast-adaptation models fall into the research areas
|
59 |
+
of transfer learning and meta learning. The dataset consists of 37,884 crowdsourced dialogues recorded between two
|
60 |
+
human users in a Wizard of Oz setup, in which one was instructed to behave like a bot, and the other a true human
|
61 |
+
user. The users are assigned a task belonging to a particular domain, for example booking a reservation at a
|
62 |
+
particular restaurant, and work together to complete the task. Our dataset spans 47 domains having 227 tasks total.
|
63 |
+
Dialogues are a minimum of 10 turns long.
|
64 |
+
|
65 |
+
### Supported Tasks and Leaderboards
|
66 |
+
|
67 |
+
This dataset supports a range of task.
|
68 |
+
- **Generative dialogue modeling** or `dialogue-modeling`: This data can be used to train task-oriented dialogue
|
69 |
+
models, specifically to develop methods to quickly simulate user responses with a small amount of data. Such fast
|
70 |
+
-adaptation models fall into the research areas of transfer learning and meta learning. The text of the dialogues
|
71 |
+
can be used to train a sequence model on the utterances.
|
72 |
+
Example of sample input/output is given in section [Data Instances](#data-instances)
|
73 |
+
|
74 |
+
|
75 |
+
|
76 |
+
### Languages
|
77 |
+
|
78 |
+
The text in the dataset is in English (`en`).
|
79 |
+
|
80 |
+
## Dataset Structure
|
81 |
+
|
82 |
+
### Data Instances
|
83 |
+
|
84 |
+
A data instance is a full multi-turn dialogue between two crowd-workers, one had the role of being a `bot`, and the other one was the `user`. Both were
|
85 |
+
given a `domain` and a `task`. Each turn has a single utterance, e.g.:
|
86 |
+
```
|
87 |
+
Domain: Ski
|
88 |
+
User Task: You want to know if there are good ski hills an
|
89 |
+
hour’s drive from your current location.
|
90 |
+
Bot Task: Tell the user that there are no ski hills in their
|
91 |
+
immediate location.
|
92 |
+
Bot: Hello how may I help you?
|
93 |
+
User: Is there any good ski hills an hour’s drive from my
|
94 |
+
current location?
|
95 |
+
Bot: I’m sorry to inform you that there are no ski hills in your
|
96 |
+
immediate location
|
97 |
+
User: Can you help me find the nearest?
|
98 |
+
Bot: Absolutely! It looks like you’re about 3 hours away from
|
99 |
+
Bear Mountain. That seems to be the closest.
|
100 |
+
User: Hmm.. sounds good
|
101 |
+
Bot: Alright! I can help you get your lift tickets now!When
|
102 |
+
will you be going?
|
103 |
+
User: Awesome! please get me a ticket for 10pax
|
104 |
+
Bot: You’ve got it. Anything else I can help you with?
|
105 |
+
User: None. Thanks again!
|
106 |
+
Bot: No problem!
|
107 |
+
```
|
108 |
+
Example of input/output for this dialog:
|
109 |
+
```
|
110 |
+
Input: dialog history = Hello how may I help you?; Is there
|
111 |
+
any good ski hills an hour’s drive from my current location?;
|
112 |
+
I’m sorry to inform you that there are no ski hills in your
|
113 |
+
immediate location
|
114 |
+
Output: user response = Can you help me find the nearest?
|
115 |
+
```
|
116 |
+
|
117 |
+
### Data Fields
|
118 |
+
|
119 |
+
Each dialogue instance has the following fields:
|
120 |
+
- `id`: a unique ID identifying the dialog.
|
121 |
+
- `user_id`: a unique ID identifying the user.
|
122 |
+
- `bot_id`: a unique ID identifying the bot.
|
123 |
+
- `domain`: a unique ID identifying the domain. Provides a mapping to tasks dataset.
|
124 |
+
- `task_id`: a unique ID identifying the task. Provides a mapping to tasks dataset.
|
125 |
+
- `turns`: the sequence of utterances alternating between `bot` and `user`, starting with a prompt from `bot`.
|
126 |
+
|
127 |
+
Each task instance has following fields:
|
128 |
+
- `task_id`: a unique ID identifying the task.
|
129 |
+
- `domain`: a unique ID identifying the domain.
|
130 |
+
- `bot_prompt`: The task specification for bot.
|
131 |
+
- `bot_role`: The domain oriented role of bot.
|
132 |
+
- `user_prompt`: The task specification for user.
|
133 |
+
- `user_role`: The domain oriented role of user.
|
134 |
+
|
135 |
+
|
136 |
+
|
137 |
+
### Data Splits
|
138 |
+
|
139 |
+
The dataset is split into a `train` and `test` split with the following sizes:
|
140 |
+
|
141 |
+
| | Training MetaLWOz | Evaluation MetaLWOz | Combined |
|
142 |
+
| ----- | ------ | ----- | ---- |
|
143 |
+
| Total Domains | 47 | 4 | 51 |
|
144 |
+
| Total Tasks | 226 | 14 | 240 |
|
145 |
+
| Total Dialogs | 37884 | 2319 | 40203 |
|
146 |
+
|
147 |
+
Below are the various statistics of the dataset:
|
148 |
+
|
149 |
+
| Statistic | Mean | Minimum | Maximum |
|
150 |
+
| ----- | ------ | ----- | ---- |
|
151 |
+
| Number of tasks per domain | 4.8 | 3 | 11 |
|
152 |
+
| Number of dialogs per domain | 806.0 | 288 | 1990 |
|
153 |
+
| Number of dialogs per task | 167.6 | 32 | 285 |
|
154 |
+
| Number of turns per dialog | 11.4 | 10 | 46 |
|
155 |
+
|
156 |
+
|
157 |
+
## Dataset Creation
|
158 |
+
|
159 |
+
### Curation Rationale
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
### Source Data
|
164 |
+
|
165 |
+
#### Initial Data Collection and Normalization
|
166 |
+
|
167 |
+
[More Information Needed]
|
168 |
+
|
169 |
+
#### Who are the source language producers?
|
170 |
+
|
171 |
+
[More Information Needed]
|
172 |
+
|
173 |
+
### Annotations
|
174 |
+
|
175 |
+
#### Annotation process
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
#### Who are the annotators?
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
### Personal and Sensitive Information
|
184 |
+
|
185 |
+
[More Information Needed]
|
186 |
+
|
187 |
+
## Considerations for Using the Data
|
188 |
+
|
189 |
+
### Social Impact of Dataset
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
### Discussion of Biases
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
### Other Known Limitations
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
|
201 |
+
## Additional Information
|
202 |
+
|
203 |
+
### Dataset Curators
|
204 |
+
The dataset v1 version is created by team of researchers from Microsoft Research (Montreal, Canada)
|
205 |
+
|
206 |
+
### Licensing Information
|
207 |
+
|
208 |
+
The dataset is released under [Microsoft Research Data License Agreement](https://msropendata-web-api.azurewebsites.net/licenses/2f933be3-284d-500b-7ea3-2aa2fd0f1bb2/view)
|
209 |
+
|
210 |
+
### Citation Information
|
211 |
+
|
212 |
+
You can cite the following for the various versions of MetaLWOz:
|
213 |
+
|
214 |
+
Version 1.0
|
215 |
+
```
|
216 |
+
@InProceedings{shalyminov2020fast,
|
217 |
+
author = {Shalyminov, Igor and Sordoni, Alessandro and Atkinson, Adam and Schulz, Hannes},
|
218 |
+
title = {Fast Domain Adaptation For Goal-Oriented Dialogue Using A Hybrid Generative-Retrieval Transformer},
|
219 |
+
booktitle = {2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
|
220 |
+
year = {2020},
|
221 |
+
month = {April},
|
222 |
+
url = {https://www.microsoft.com/en-us/research/publication/fast-domain-adaptation-for-goal-oriented-dialogue-using-a
|
223 |
+
-hybrid-generative-retrieval-transformer/},
|
224 |
+
}
|
225 |
+
```
|
dataset_infos.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dialogues": {"description": "MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models. We introduce the Meta-Learning Wizard of Oz (MetaLWOz) dialogue dataset for developing fast adaptation methods for conversation models. This data can be used to train task-oriented dialogue models, specifically to develop methods to quickly simulate user responses with a small amount of data. Such fast-adaptation models fall into the research areas of transfer learning and meta learning. The dataset consists of 37,884 crowdsourced dialogues recorded between two human users in a Wizard of Oz setup, in which one was instructed to behave like a bot, and the other a true human user. The users are assigned a task belonging to a particular domain, for example booking a reservation at a particular restaurant, and work together to complete the task. Our dataset spans 47 domains having 227 tasks total. Dialogues are a minimum of 10 turns long.\n", "citation": "@InProceedings{shalyminov2020fast,\nauthor = {Shalyminov, Igor and Sordoni, Alessandro and Atkinson, Adam and Schulz, Hannes},\ntitle = {Fast Domain Adaptation For Goal-Oriented Dialogue Using A Hybrid Generative-Retrieval Transformer},\nbooktitle = {2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},\nyear = {2020},\nmonth = {April},\nurl = {https://www.microsoft.com/en-us/research/publication/fast-domain-adaptation-for-goal-oriented-dialogue-using-a-hybrid-generative-retrieval-transformer/},\n}\n", "homepage": "https://www.microsoft.com/en-us/research/project/metalwoz/", "license": "Microsoft Research Data License Agreement", "features": {"id": {"dtype": "string", "id": null, "_type": "Value"}, "user_id": {"dtype": "string", "id": null, "_type": "Value"}, "bot_id": {"dtype": "string", "id": null, "_type": "Value"}, "domain": {"dtype": "string", "id": null, "_type": "Value"}, "task_id": {"dtype": "string", "id": null, "_type": "Value"}, "turns": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}}, "post_processed": null, "supervised_keys": null, "builder_name": "meta_woz", "config_name": "dialogues", "version": {"version_str": "0.0.0", "description": null, "major": 0, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 19999218, "num_examples": 37884, "dataset_name": "meta_woz"}, "test": {"name": "test", "num_bytes": 1284287, "num_examples": 2319, "dataset_name": "meta_woz"}}, "download_checksums": {"https://download.microsoft.com/download/E/B/8/EB84CB1A-D57D-455F-B905-3ABDE80404E5/metalwoz-v1.zip": {"num_bytes": 5639228, "checksum": "2a2ae3b25760aa2725e70bc6480562fa5d720c9689a508d28417631496d6764f"}, "https://download.microsoft.com/download/0/c/4/0c4a8893-cbf9-4a43-a44a-09bab9539234/metalwoz-test-v1.zip": {"num_bytes": 2990635, "checksum": "6722d1d9ec05334dd801972767ae3bdefcd15f71bf73fea1d098f214a96a7c6c"}}, "download_size": 8629863, "post_processing_size": null, "dataset_size": 21283505, "size_in_bytes": 29913368}, "tasks": {"description": "MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models. We introduce the Meta-Learning Wizard of Oz (MetaLWOz) dialogue dataset for developing fast adaptation methods for conversation models. This data can be used to train task-oriented dialogue models, specifically to develop methods to quickly simulate user responses with a small amount of data. Such fast-adaptation models fall into the research areas of transfer learning and meta learning. The dataset consists of 37,884 crowdsourced dialogues recorded between two human users in a Wizard of Oz setup, in which one was instructed to behave like a bot, and the other a true human user. The users are assigned a task belonging to a particular domain, for example booking a reservation at a particular restaurant, and work together to complete the task. Our dataset spans 47 domains having 227 tasks total. Dialogues are a minimum of 10 turns long.\n", "citation": "@InProceedings{shalyminov2020fast,\nauthor = {Shalyminov, Igor and Sordoni, Alessandro and Atkinson, Adam and Schulz, Hannes},\ntitle = {Fast Domain Adaptation For Goal-Oriented Dialogue Using A Hybrid Generative-Retrieval Transformer},\nbooktitle = {2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},\nyear = {2020},\nmonth = {April},\nurl = {https://www.microsoft.com/en-us/research/publication/fast-domain-adaptation-for-goal-oriented-dialogue-using-a-hybrid-generative-retrieval-transformer/},\n}\n", "homepage": "https://www.microsoft.com/en-us/research/project/metalwoz/", "license": "Microsoft Research Data License Agreement", "features": {"task_id": {"dtype": "string", "id": null, "_type": "Value"}, "domain": {"dtype": "string", "id": null, "_type": "Value"}, "bot_prompt": {"dtype": "string", "id": null, "_type": "Value"}, "bot_role": {"dtype": "string", "id": null, "_type": "Value"}, "user_prompt": {"dtype": "string", "id": null, "_type": "Value"}, "user_role": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "meta_woz", "config_name": "tasks", "version": "0.0.0", "splits": {"train": {"name": "train", "num_bytes": 73768, "num_examples": 227, "dataset_name": "meta_woz"}, "test": {"name": "test", "num_bytes": 4351, "num_examples": 14, "dataset_name": "meta_woz"}}, "download_checksums": {"https://download.microsoft.com/download/E/B/8/EB84CB1A-D57D-455F-B905-3ABDE80404E5/metalwoz-v1.zip": {"num_bytes": 5639228, "checksum": "2a2ae3b25760aa2725e70bc6480562fa5d720c9689a508d28417631496d6764f"}, "https://download.microsoft.com/download/0/c/4/0c4a8893-cbf9-4a43-a44a-09bab9539234/metalwoz-test-v1.zip": {"num_bytes": 2990635, "checksum": "6722d1d9ec05334dd801972767ae3bdefcd15f71bf73fea1d098f214a96a7c6c"}}, "download_size": 8629863, "post_processing_size": null, "dataset_size": 78119, "size_in_bytes": 8707982}}
|
dummy/dialogues/0.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cb8adb221ac474cbfcef27a9daf3e34d3f8403c617d37202db077723dae09124
|
3 |
+
size 16460
|
dummy/tasks/0.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:60faf495c7c6ae2ed2807ec091886b299a2a7a5168becc61cd8b43fb79c1c44e
|
3 |
+
size 8084
|
meta_woz.py
ADDED
@@ -0,0 +1,154 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
"""MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models"""
|
16 |
+
|
17 |
+
from __future__ import absolute_import, division, print_function
|
18 |
+
|
19 |
+
import json
|
20 |
+
import os
|
21 |
+
|
22 |
+
import datasets
|
23 |
+
|
24 |
+
|
25 |
+
_CITATION = """\
|
26 |
+
@InProceedings{shalyminov2020fast,
|
27 |
+
author = {Shalyminov, Igor and Sordoni, Alessandro and Atkinson, Adam and Schulz, Hannes},
|
28 |
+
title = {Fast Domain Adaptation For Goal-Oriented Dialogue Using A Hybrid Generative-Retrieval Transformer},
|
29 |
+
booktitle = {2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
|
30 |
+
year = {2020},
|
31 |
+
month = {April},
|
32 |
+
url = {https://www.microsoft.com/en-us/research/publication/fast-domain-adaptation-for-goal-oriented-dialogue-using-a
|
33 |
+
-hybrid-generative-retrieval-transformer/},
|
34 |
+
}
|
35 |
+
"""
|
36 |
+
|
37 |
+
_DESCRIPTION = """\
|
38 |
+
MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models. \
|
39 |
+
We introduce the Meta-Learning Wizard of Oz (MetaLWOz) dialogue dataset for developing fast adaptation methods for \
|
40 |
+
conversation models. This data can be used to train task-oriented dialogue models, specifically to develop methods to \
|
41 |
+
quickly simulate user responses with a small amount of data. Such fast-adaptation models fall into the research areas \
|
42 |
+
of transfer learning and meta learning. The dataset consists of 37,884 crowdsourced dialogues recorded between two \
|
43 |
+
human users in a Wizard of Oz setup, in which one was instructed to behave like a bot, and the other a true human \
|
44 |
+
user. The users are assigned a task belonging to a particular domain, for example booking a reservation at a \
|
45 |
+
particular restaurant, and work together to complete the task. Our dataset spans 47 domains having 227 tasks total. \
|
46 |
+
Dialogues are a minimum of 10 turns long.
|
47 |
+
"""
|
48 |
+
|
49 |
+
_HOMEPAGE = "https://www.microsoft.com/en-us/research/project/metalwoz/"
|
50 |
+
|
51 |
+
_LICENSE = "Microsoft Research Data License Agreement"
|
52 |
+
|
53 |
+
_URLs = {
|
54 |
+
"train": "https://download.microsoft.com/download/E/B/8/EB84CB1A-D57D-455F-B905-3ABDE80404E5/metalwoz-v1.zip",
|
55 |
+
"test": "https://download.microsoft.com/download/0/c/4/0c4a8893-cbf9-4a43-a44a-09bab9539234/metalwoz-test-v1.zip",
|
56 |
+
}
|
57 |
+
|
58 |
+
|
59 |
+
class MetaWoz(datasets.GeneratorBasedBuilder):
|
60 |
+
VERSION = datasets.Version("1.0.0")
|
61 |
+
|
62 |
+
BUILDER_CONFIGS = [
|
63 |
+
datasets.BuilderConfig(name="dialogues", description="The dataset of dialogues from various domains."),
|
64 |
+
datasets.BuilderConfig(
|
65 |
+
name="tasks", description="The metadata for tasks corresponding to dialogues from " "various domains."
|
66 |
+
),
|
67 |
+
]
|
68 |
+
|
69 |
+
DEFAULT_CONFIG_NAME = "dialogues"
|
70 |
+
|
71 |
+
def _info(self):
|
72 |
+
if self.config.name == "tasks":
|
73 |
+
features = datasets.Features(
|
74 |
+
{
|
75 |
+
"task_id": datasets.Value("string"),
|
76 |
+
"domain": datasets.Value("string"),
|
77 |
+
"bot_prompt": datasets.Value("string"),
|
78 |
+
"bot_role": datasets.Value("string"),
|
79 |
+
"user_prompt": datasets.Value("string"),
|
80 |
+
"user_role": datasets.Value("string"),
|
81 |
+
}
|
82 |
+
)
|
83 |
+
else:
|
84 |
+
features = datasets.Features(
|
85 |
+
{
|
86 |
+
"id": datasets.Value("string"),
|
87 |
+
"user_id": datasets.Value("string"),
|
88 |
+
"bot_id": datasets.Value("string"),
|
89 |
+
"domain": datasets.Value("string"),
|
90 |
+
"task_id": datasets.Value("string"),
|
91 |
+
"turns": datasets.Sequence(datasets.Value("string")),
|
92 |
+
}
|
93 |
+
)
|
94 |
+
return datasets.DatasetInfo(
|
95 |
+
description=_DESCRIPTION,
|
96 |
+
features=features,
|
97 |
+
supervised_keys=None,
|
98 |
+
homepage=_HOMEPAGE,
|
99 |
+
license=_LICENSE,
|
100 |
+
citation=_CITATION,
|
101 |
+
)
|
102 |
+
|
103 |
+
def _split_generators(self, dl_manager):
|
104 |
+
"""Returns SplitGenerators."""
|
105 |
+
data_dir = dl_manager.download_and_extract(_URLs)
|
106 |
+
data_dir["test"] = dl_manager.extract(os.path.join(data_dir["test"], "dstc8_metalwoz_heldout.zip"))
|
107 |
+
|
108 |
+
return [
|
109 |
+
datasets.SplitGenerator(
|
110 |
+
name=datasets.Split.TRAIN,
|
111 |
+
# These kwargs will be passed to _generate_examples
|
112 |
+
gen_kwargs={"data_dir": data_dir["train"]},
|
113 |
+
),
|
114 |
+
datasets.SplitGenerator(
|
115 |
+
name=datasets.Split.TEST,
|
116 |
+
# These kwargs will be passed to _generate_examples
|
117 |
+
gen_kwargs={"data_dir": data_dir["test"]},
|
118 |
+
),
|
119 |
+
]
|
120 |
+
|
121 |
+
def _generate_examples(self, data_dir):
|
122 |
+
""" Yields examples. """
|
123 |
+
if self.config.name == "tasks":
|
124 |
+
filepath = os.path.join(data_dir, "tasks.txt")
|
125 |
+
with open(filepath, encoding="utf-8") as f:
|
126 |
+
for id_, row in enumerate(f):
|
127 |
+
data = json.loads(row)
|
128 |
+
yield id_, {
|
129 |
+
"task_id": data["task_id"],
|
130 |
+
"domain": data["domain"],
|
131 |
+
"bot_prompt": data["bot_prompt"],
|
132 |
+
"bot_role": data["bot_role"],
|
133 |
+
"user_prompt": data["user_prompt"],
|
134 |
+
"user_role": data["user_role"],
|
135 |
+
}
|
136 |
+
else:
|
137 |
+
id_ = -1
|
138 |
+
base_path = os.path.join(data_dir, "dialogues")
|
139 |
+
file_list = sorted(
|
140 |
+
[os.path.join(base_path, file) for file in os.listdir(base_path) if file.endswith(".txt")]
|
141 |
+
)
|
142 |
+
for filepath in file_list:
|
143 |
+
with open(filepath, encoding="utf-8") as f:
|
144 |
+
for row in f:
|
145 |
+
id_ += 1
|
146 |
+
data = json.loads(row)
|
147 |
+
yield id_, {
|
148 |
+
"id": data["id"],
|
149 |
+
"user_id": data["user_id"],
|
150 |
+
"bot_id": data["bot_id"],
|
151 |
+
"domain": data["domain"],
|
152 |
+
"task_id": data["task_id"],
|
153 |
+
"turns": data["turns"],
|
154 |
+
}
|