Datasets:
Commit
•
95f0b73
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 +252 -0
- dataset_infos.json +1 -0
- dummy/flights/1.0.0/dummy_data.zip +3 -0
- dummy/food-ordering/1.0.0/dummy_data.zip +3 -0
- dummy/hotels/1.0.0/dummy_data.zip +3 -0
- dummy/movies/1.0.0/dummy_data.zip +3 -0
- dummy/music/1.0.0/dummy_data.zip +3 -0
- dummy/restaurant-search/1.0.0/dummy_data.zip +3 -0
- dummy/sports/1.0.0/dummy_data.zip +3 -0
- taskmaster2.py +126 -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 +1,252 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
annotations_creators:
|
3 |
+
- crowdsourced
|
4 |
+
language_creators:
|
5 |
+
- crowdsourced
|
6 |
+
languages:
|
7 |
+
- en
|
8 |
+
licenses:
|
9 |
+
- cc-by-4-0
|
10 |
+
multilinguality:
|
11 |
+
- monolingual
|
12 |
+
size_categories:
|
13 |
+
- 1K<n<10K
|
14 |
+
source_datasets:
|
15 |
+
- original
|
16 |
+
task_categories:
|
17 |
+
- sequence-modeling
|
18 |
+
task_ids:
|
19 |
+
- dialogue-modeling
|
20 |
+
---
|
21 |
+
|
22 |
+
# Dataset Card Creation Guide
|
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 |
+
- **Homepage:** [Taskmaster-1](https://research.google/tools/datasets/taskmaster-1/)
|
50 |
+
- **Repository:** [GitHub](https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020)
|
51 |
+
- **Paper:** [Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset](https://arxiv.org/abs/1909.05358)
|
52 |
+
- **Leaderboard:** N/A
|
53 |
+
- **Point of Contact:** [Taskmaster Googlegroup](taskmaster-datasets@googlegroups.com)
|
54 |
+
|
55 |
+
### Dataset Summary
|
56 |
+
|
57 |
+
Taskmaster is dataset for goal oriented conversations. The Taskmaster-2 dataset consists of 17,289 dialogs
|
58 |
+
in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports.
|
59 |
+
Unlike Taskmaster-1, which includes both written "self-dialogs" and spoken two-person dialogs,
|
60 |
+
Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is
|
61 |
+
almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs.
|
62 |
+
All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced
|
63 |
+
workers played the role of a 'user' and trained call center operators played the role of the 'assistant'.
|
64 |
+
In this way, users were led to believe they were interacting with an automated system that “spoke”
|
65 |
+
using text-to-speech (TTS) even though it was in fact a human behind the scenes.
|
66 |
+
As a result, users could express themselves however they chose in the context of an automated interface.
|
67 |
+
|
68 |
+
### Supported Tasks and Leaderboards
|
69 |
+
|
70 |
+
[More Information Needed]
|
71 |
+
|
72 |
+
### Languages
|
73 |
+
|
74 |
+
The dataset is in English language.
|
75 |
+
|
76 |
+
## Dataset Structure
|
77 |
+
|
78 |
+
### Data Instances
|
79 |
+
|
80 |
+
A typical example looks like this
|
81 |
+
|
82 |
+
```
|
83 |
+
{
|
84 |
+
"conversation_id": "dlg-0047a087-6a3c-4f27-b0e6-268f53a2e013",
|
85 |
+
"instruction_id": "flight-6",
|
86 |
+
"utterances": [
|
87 |
+
{
|
88 |
+
"index": 0,
|
89 |
+
"segments": [],
|
90 |
+
"speaker": "USER",
|
91 |
+
"text": "Hi, I'm looking for a flight. I need to visit a friend."
|
92 |
+
},
|
93 |
+
{
|
94 |
+
"index": 1,
|
95 |
+
"segments": [],
|
96 |
+
"speaker": "ASSISTANT",
|
97 |
+
"text": "Hello, how can I help you?"
|
98 |
+
},
|
99 |
+
{
|
100 |
+
"index": 2,
|
101 |
+
"segments": [],
|
102 |
+
"speaker": "ASSISTANT",
|
103 |
+
"text": "Sure, I can help you with that."
|
104 |
+
},
|
105 |
+
{
|
106 |
+
"index": 3,
|
107 |
+
"segments": [],
|
108 |
+
"speaker": "ASSISTANT",
|
109 |
+
"text": "On what dates?"
|
110 |
+
},
|
111 |
+
{
|
112 |
+
"index": 4,
|
113 |
+
"segments": [
|
114 |
+
{
|
115 |
+
"annotations": [
|
116 |
+
{
|
117 |
+
"name": "flight_search.date.depart_origin"
|
118 |
+
}
|
119 |
+
],
|
120 |
+
"end_index": 37,
|
121 |
+
"start_index": 27,
|
122 |
+
"text": "March 20th"
|
123 |
+
},
|
124 |
+
{
|
125 |
+
"annotations": [
|
126 |
+
{
|
127 |
+
"name": "flight_search.date.return"
|
128 |
+
}
|
129 |
+
],
|
130 |
+
"end_index": 45,
|
131 |
+
"start_index": 41,
|
132 |
+
"text": "22nd"
|
133 |
+
}
|
134 |
+
],
|
135 |
+
"speaker": "USER",
|
136 |
+
"text": "I'm looking to travel from March 20th to 22nd."
|
137 |
+
}
|
138 |
+
]
|
139 |
+
}
|
140 |
+
```
|
141 |
+
|
142 |
+
### Data Fields
|
143 |
+
|
144 |
+
Each conversation in the data file has the following structure:
|
145 |
+
|
146 |
+
- `conversation_id`: A universally unique identifier with the prefix 'dlg-'. The ID has no meaning.
|
147 |
+
- `utterances`: A list of utterances that make up the conversation.
|
148 |
+
- `instruction_id`: A reference to the file(s) containing the user (and, if applicable, agent) instructions for this conversation.
|
149 |
+
|
150 |
+
Each utterance has the following fields:
|
151 |
+
|
152 |
+
- `index`: A 0-based index indicating the order of the utterances in the conversation.
|
153 |
+
- `speaker`: Either USER or ASSISTANT, indicating which role generated this utterance.
|
154 |
+
- `text`: The raw text of the utterance. In case of self dialogs (one_person_dialogs), this is written by the crowdsourced worker. In case of the WOz dialogs, 'ASSISTANT' turns are written and 'USER' turns are transcribed from the spoken recordings of crowdsourced workers.
|
155 |
+
- `segments`: A list of various text spans with semantic annotations.
|
156 |
+
|
157 |
+
Each segment has the following fields:
|
158 |
+
|
159 |
+
- `start_index`: The position of the start of the annotation in the utterance text.
|
160 |
+
- `end_index`: The position of the end of the annotation in the utterance text.
|
161 |
+
- `text`: The raw text that has been annotated.
|
162 |
+
- `annotations`: A list of annotation details for this segment.
|
163 |
+
|
164 |
+
Each annotation has a single field:
|
165 |
+
|
166 |
+
- `name`: The annotation name.
|
167 |
+
|
168 |
+
|
169 |
+
|
170 |
+
### Data Splits
|
171 |
+
|
172 |
+
There are no deafults splits for all the config. The below table lists the number of examples in each config.
|
173 |
+
|
174 |
+
| Config | Train |
|
175 |
+
|-------------------|--------|
|
176 |
+
| flights | 2481 |
|
177 |
+
| food-orderings | 1050 |
|
178 |
+
| hotels | 2355 |
|
179 |
+
| movies | 3047 |
|
180 |
+
| music | 1602 |
|
181 |
+
| restaurant-search | 3276 |
|
182 |
+
| sports | 3478 |
|
183 |
+
|
184 |
+
|
185 |
+
## Dataset Creation
|
186 |
+
|
187 |
+
### Curation Rationale
|
188 |
+
|
189 |
+
[More Information Needed]
|
190 |
+
|
191 |
+
### Source Data
|
192 |
+
|
193 |
+
[More Information Needed]
|
194 |
+
|
195 |
+
#### Initial Data Collection and Normalization
|
196 |
+
|
197 |
+
[More Information Needed]
|
198 |
+
|
199 |
+
#### Who are the source language producers?
|
200 |
+
|
201 |
+
[More Information Needed]
|
202 |
+
|
203 |
+
### Annotations
|
204 |
+
|
205 |
+
[More Information Needed]
|
206 |
+
|
207 |
+
#### Annotation process
|
208 |
+
|
209 |
+
[More Information Needed]
|
210 |
+
|
211 |
+
#### Who are the annotators?
|
212 |
+
|
213 |
+
[More Information Needed]
|
214 |
+
|
215 |
+
### Personal and Sensitive Information
|
216 |
+
|
217 |
+
[More Information Needed]
|
218 |
+
|
219 |
+
## Considerations for Using the Data
|
220 |
+
|
221 |
+
### Social Impact of Dataset
|
222 |
+
|
223 |
+
[More Information Needed]
|
224 |
+
|
225 |
+
### Discussion of Biases
|
226 |
+
|
227 |
+
[More Information Needed]
|
228 |
+
|
229 |
+
### Other Known Limitations
|
230 |
+
|
231 |
+
[More Information Needed]
|
232 |
+
|
233 |
+
## Additional Information
|
234 |
+
|
235 |
+
### Dataset Curators
|
236 |
+
|
237 |
+
[More Information Needed]
|
238 |
+
|
239 |
+
### Licensing Information
|
240 |
+
|
241 |
+
The dataset is licensed under `Creative Commons Attribution 4.0 License`
|
242 |
+
|
243 |
+
### Citation Information
|
244 |
+
|
245 |
+
[More Information Needed]
|
246 |
+
```
|
247 |
+
@inproceedings{48484,
|
248 |
+
title = {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},
|
249 |
+
author = {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},
|
250 |
+
year = {2019}
|
251 |
+
}
|
252 |
+
```
|
dataset_infos.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"flights": {"description": "Taskmaster is dataset for goal oriented conversations. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "flights", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 7073487, "num_examples": 2481, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/flights.json": {"num_bytes": 23029880, "checksum": "86b37b5ae25f530fd18ced78800d30c3b54f7b34bb208ecb51842718f04e760b"}}, "download_size": 23029880, "post_processing_size": null, "dataset_size": 7073487, "size_in_bytes": 30103367}, "food-ordering": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "food-ordering", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1734825, "num_examples": 1050, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/food-ordering.json": {"num_bytes": 5376675, "checksum": "0a042e566a816a5d0abebe6f7e8cfd6abaa89729ffc42f433d327df7342b12f8"}}, "download_size": 5376675, "post_processing_size": null, "dataset_size": 1734825, "size_in_bytes": 7111500}, "hotels": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "hotels", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 7436667, "num_examples": 2357, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/hotels.json": {"num_bytes": 22507266, "checksum": "975b0242f1e37ea1ab94ccedd7e0d6ee5831599d5df1f16143e71110d6c6006a"}}, "download_size": 22507266, "post_processing_size": null, "dataset_size": 7436667, "size_in_bytes": 29943933}, "movies": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "movies", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 7112301, "num_examples": 3056, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/movies.json": {"num_bytes": 21189893, "checksum": "6f67c9a1f04abc111186e5bcfbe3050be01d0737fd6422901402715bc1f3dd0d"}}, "download_size": 21189893, "post_processing_size": null, "dataset_size": 7112301, "size_in_bytes": 28302194}, "music": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "music", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 2814030, "num_examples": 1603, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/music.json": {"num_bytes": 8981720, "checksum": "e5db60d6576fa010bef87a70a8b371d293d48cde8524c1d3ed7c3022f079d95d"}}, "download_size": 8981720, "post_processing_size": null, "dataset_size": 2814030, "size_in_bytes": 11795750}, "restaurant-search": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "restaurant-search", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 7341998, "num_examples": 3276, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/restaurant-search.json": {"num_bytes": 21472680, "checksum": "fb9735f89e7ebc7c877f976da4c30391af6a6277991b597c0755564657ff8f47"}}, "download_size": 21472680, "post_processing_size": null, "dataset_size": 7341998, "size_in_bytes": 28814678}, "sports": {"description": "Taskmaster is dataset for goal oriented conversationas. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written \"self-dialogs\" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that \u201cspoke\u201d using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.\n", "citation": "@inproceedings{48484,\ntitle\t= {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},\nauthor\t= {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},\nyear\t= {2019}\n}\n", "homepage": "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020", "license": "", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "instruction_id": {"dtype": "string", "id": null, "_type": "Value"}, "utterances": [{"index": {"dtype": "int32", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "segments": [{"start_index": {"dtype": "int32", "id": null, "_type": "Value"}, "end_index": {"dtype": "int32", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "annotations": [{"name": {"dtype": "string", "id": null, "_type": "Value"}}]}]}]}, "post_processed": null, "supervised_keys": null, "builder_name": "taskmaster2", "config_name": "sports", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 5738818, "num_examples": 3481, "dataset_name": "taskmaster2"}}, "download_checksums": {"https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data/sports.json": {"num_bytes": 19549440, "checksum": "8191531bfa5a8426b1508c396ab9886a19c7c620b443c436ec10d8d4708d0eac"}}, "download_size": 19549440, "post_processing_size": null, "dataset_size": 5738818, "size_in_bytes": 25288258}}
|
dummy/flights/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6601271877fecec2624389bd5b84cf68ac10cd90e504aeb1280c9ef99dbd4d02
|
3 |
+
size 4615
|
dummy/food-ordering/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:90c8e5623884617f1f85c87100333a97094c59dcb1e25fbcdf02e51e4237af24
|
3 |
+
size 3261
|
dummy/hotels/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f0e99c216f227d59d198487f079eafd7d0409b0991da7c20ce63e996196aa3fb
|
3 |
+
size 5368
|
dummy/movies/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:edd0aee5bcbdc81165eaa8237b382e0bed14483b9866ff071dbc2b2a7855df28
|
3 |
+
size 2739
|
dummy/music/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:28c1d6ea95bf798628661ca2e38ca9bde95bd4db06210b55dfd051d603efeff6
|
3 |
+
size 2542
|
dummy/restaurant-search/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f85af3d8124dcead610a4288b1ef76426838627448cd6c9bc3bb35ff29d5280c
|
3 |
+
size 3138
|
dummy/sports/1.0.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:109134033b5dd0ad454e04bbf8065c4d66a1d3918809b428d34d8d977955f27a
|
3 |
+
size 2710
|
taskmaster2.py
ADDED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
"""Taskmaster: A dataset for goal oriented conversations."""
|
16 |
+
|
17 |
+
from __future__ import absolute_import, division, print_function
|
18 |
+
|
19 |
+
import json
|
20 |
+
|
21 |
+
import datasets
|
22 |
+
|
23 |
+
|
24 |
+
_CITATION = """\
|
25 |
+
@inproceedings{48484,
|
26 |
+
title = {Taskmaster-1: Toward a Realistic and Diverse Dialog Dataset},
|
27 |
+
author = {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},
|
28 |
+
year = {2019}
|
29 |
+
}
|
30 |
+
"""
|
31 |
+
|
32 |
+
_DESCRIPTION = """\
|
33 |
+
Taskmaster is dataset for goal oriented conversations. The Taskmaster-2 dataset consists of 17,289 dialogs \
|
34 |
+
in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. \
|
35 |
+
Unlike Taskmaster-1, which includes both written "self-dialogs" and spoken two-person dialogs, \
|
36 |
+
Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is \
|
37 |
+
almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. \
|
38 |
+
All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced \
|
39 |
+
workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. \
|
40 |
+
In this way, users were led to believe they were interacting with an automated system that “spoke” \
|
41 |
+
using text-to-speech (TTS) even though it was in fact a human behind the scenes. \
|
42 |
+
As a result, users could express themselves however they chose in the context of an automated interface.
|
43 |
+
"""
|
44 |
+
|
45 |
+
_HOMEPAGE = "https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020"
|
46 |
+
|
47 |
+
_BASE_URL = "https://raw.githubusercontent.com/google-research-datasets/Taskmaster/master/TM-2-2020/data"
|
48 |
+
|
49 |
+
|
50 |
+
class Taskmaster2(datasets.GeneratorBasedBuilder):
|
51 |
+
"""Taskmaster: A dataset for goal oriented conversations."""
|
52 |
+
|
53 |
+
VERSION = datasets.Version("1.0.0")
|
54 |
+
BUILDER_CONFIGS = [
|
55 |
+
datasets.BuilderConfig(
|
56 |
+
name="flights", version=datasets.Version("1.0.0"), description="Taskmaster-2 flights domain."
|
57 |
+
),
|
58 |
+
datasets.BuilderConfig(
|
59 |
+
name="food-ordering", version=datasets.Version("1.0.0"), description="Taskmaster-2 food-ordering domain"
|
60 |
+
),
|
61 |
+
datasets.BuilderConfig(
|
62 |
+
name="hotels", version=datasets.Version("1.0.0"), description="Taskmaster-2 hotel domain"
|
63 |
+
),
|
64 |
+
datasets.BuilderConfig(
|
65 |
+
name="movies", version=datasets.Version("1.0.0"), description="Taskmaster-2 movies domain"
|
66 |
+
),
|
67 |
+
datasets.BuilderConfig(
|
68 |
+
name="music", version=datasets.Version("1.0.0"), description="Taskmaster-2 music domain"
|
69 |
+
),
|
70 |
+
datasets.BuilderConfig(
|
71 |
+
name="restaurant-search",
|
72 |
+
version=datasets.Version("1.0.0"),
|
73 |
+
description="Taskmaster-2 restaurant-search domain",
|
74 |
+
),
|
75 |
+
datasets.BuilderConfig(
|
76 |
+
name="sports", version=datasets.Version("1.0.0"), description="Taskmaster-2 sports domain"
|
77 |
+
),
|
78 |
+
]
|
79 |
+
|
80 |
+
def _info(self):
|
81 |
+
features = {
|
82 |
+
"conversation_id": datasets.Value("string"),
|
83 |
+
"instruction_id": datasets.Value("string"),
|
84 |
+
"utterances": [
|
85 |
+
{
|
86 |
+
"index": datasets.Value("int32"),
|
87 |
+
"speaker": datasets.Value("string"),
|
88 |
+
"text": datasets.Value("string"),
|
89 |
+
"segments": [
|
90 |
+
{
|
91 |
+
"start_index": datasets.Value("int32"),
|
92 |
+
"end_index": datasets.Value("int32"),
|
93 |
+
"text": datasets.Value("string"),
|
94 |
+
"annotations": [{"name": datasets.Value("string")}],
|
95 |
+
}
|
96 |
+
],
|
97 |
+
}
|
98 |
+
],
|
99 |
+
}
|
100 |
+
return datasets.DatasetInfo(
|
101 |
+
description=_DESCRIPTION,
|
102 |
+
features=datasets.Features(features),
|
103 |
+
supervised_keys=None,
|
104 |
+
homepage=_HOMEPAGE,
|
105 |
+
citation=_CITATION,
|
106 |
+
)
|
107 |
+
|
108 |
+
def _split_generators(self, dl_manager):
|
109 |
+
url = f"{_BASE_URL}/{self.config.name}.json"
|
110 |
+
dialogs_file = dl_manager.download(url)
|
111 |
+
return [
|
112 |
+
datasets.SplitGenerator(
|
113 |
+
name=datasets.Split.TRAIN,
|
114 |
+
gen_kwargs={"filepath": dialogs_file},
|
115 |
+
),
|
116 |
+
]
|
117 |
+
|
118 |
+
def _generate_examples(self, filepath):
|
119 |
+
with open(filepath, encoding="utf-8") as f:
|
120 |
+
dialogs = json.load(f)
|
121 |
+
for dialog in dialogs:
|
122 |
+
utterances = dialog["utterances"]
|
123 |
+
for utterance in utterances:
|
124 |
+
if "segments" not in utterance:
|
125 |
+
utterance["segments"] = []
|
126 |
+
yield dialog["conversation_id"], dialog
|