Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -20,7 +20,9 @@ task_ids:
|
|
20 |
- dialogue-modeling
|
21 |
---
|
22 |
# Dataset Card for RiSAWOZ
|
|
|
23 |
## Table of Contents
|
|
|
24 |
- [Dataset Card for RiSAWOZ](#dataset-card-for-risawoz)
|
25 |
- [Table of Contents](#table-of-contents)
|
26 |
- [Dataset Description](#dataset-description)
|
@@ -45,25 +47,37 @@ task_ids:
|
|
45 |
- [Licensing Information](#licensing-information)
|
46 |
- [Citation Information](#citation-information)
|
47 |
- [Contributions](#contributions)
|
|
|
48 |
## Dataset Description
|
49 |
-
|
50 |
-
- **
|
51 |
-
- **
|
|
|
52 |
- **Point of Contact:** Deyi Xiong (dyxiong@tju.edu.cn)
|
|
|
53 |
### Dataset Summary
|
|
|
54 |
RiSAWOZ contains 11.2K human-to-human (H2H) multiturn semantically annotated dialogues, with more than 150K utterances spanning over 12 domains, which is larger than all previous annotated H2H conversational datasets. Both single- and multi-domain dialogues are constructed, accounting for 65% and 35%, respectively.
|
|
|
55 |
### Supported Tasks and Leaderboards
|
|
|
56 |
- Natural Language Understanding
|
57 |
- Dialogue State Tracking
|
58 |
- Dialogue Context-to-Text Generation
|
59 |
- Coreference Resolution
|
60 |
- Unified Generative Ellipsis and Coreference Resolution
|
|
|
61 |
### Languages
|
62 |
-
|
|
|
|
|
63 |
## Dataset Structure
|
|
|
64 |
### Data Instances
|
|
|
65 |
A JSON formatted example of a typical instance in RiSAWOZ dataset:
|
66 |
-
|
|
|
67 |
{
|
68 |
"dialogue_id": "attraction_goal_4-63###6177",
|
69 |
"goal": "attraction_goal_4-63: 你是苏州人,但不怎么出去玩。你朋友来苏州找你,你准备带他逛逛“水乡古镇”,你希望客服给你推荐个消费水平“中等”的地方。然后你要问清楚这地方“是否地铁直达”、“特点”、“门票价格”这些信息。最后,你要感谢客服的帮助,然后说再见。",
|
@@ -577,66 +591,93 @@ A JSON formatted example of a typical instance in RiSAWOZ dataset:
|
|
577 |
]
|
578 |
}
|
579 |
```
|
|
|
580 |
### Data Fields
|
|
|
581 |
The data fields are the same among all splits.
|
|
|
582 |
- dialogue_id (string): dialogue ID
|
583 |
- goal (string): natural language descriptions of the user goal
|
584 |
- domains (list of strings): domains mentioned in current dialogue session
|
585 |
- dialogue (list of dicts): dialog turns and corresponding annotations. Each turn includes:
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
|
590 |
-
|
591 |
-
|
592 |
-
|
593 |
-
|
594 |
-
|
595 |
-
|
596 |
-
|
597 |
-
|
598 |
-
|
|
|
599 |
### Data Splits
|
|
|
600 |
| | train | validation | test |
|
601 |
| --------- | -----: | ---------: | ----: |
|
602 |
| #dialogues | 10000 | 600 | 600 |
|
603 |
| #turns | 134580 | 8116 | 9286 |
|
604 |
| Avg. turns | 13.5 | 13.5 | 15.5 |
|
|
|
605 |
## Dataset Creation
|
|
|
606 |
### Curation Rationale
|
|
|
607 |
Gather human-to-human dialog in Chinese.
|
|
|
608 |
### Communicative Goal
|
|
|
609 |
Generate system response given dialogue context across multiple domains.
|
|
|
610 |
### Annotations
|
|
|
611 |
#### Annotation process
|
|
|
612 |
- Does the dataset have additional annotations for each instance?
|
613 |
-
|
614 |
- What is the number of raters
|
615 |
-
|
616 |
- Describe the qualifications required of an annotator.
|
617 |
-
|
618 |
- How many annotators saw each training example?
|
619 |
-
|
620 |
- How many annotators saw each test example?
|
621 |
-
|
|
|
622 |
### Personal and Sensitive Information
|
|
|
623 |
The slots and values as well as utterances do not contain any personal information.
|
|
|
624 |
## Considerations for Using the Data
|
|
|
625 |
### Social Impact of Dataset
|
|
|
626 |
RiSAWOZ is the first large-scale multi-domain Chinese Wizard-of-Oz dataset with rich semantic annotations.
|
|
|
627 |
## Additional Information
|
|
|
628 |
### Dataset Curators
|
|
|
629 |
- Jun Quan (Soochow University, Suzhou, China)
|
630 |
- Shian Zhang (Soochow University, Suzhou, China)
|
631 |
- Qian Cao (Soochow University, Suzhou, China)
|
632 |
- Zizhong Li (Tianjin University, Tianjin, China)
|
633 |
- Deyi Xiong (Tianjin University, Tianjin, China)
|
|
|
634 |
### Funding Information
|
|
|
635 |
the National Key Research and Development Project
|
|
|
636 |
### Licensing Information
|
|
|
637 |
cc-by-4.0: Creative Commons Attribution 4.0 International
|
|
|
638 |
### Citation Information
|
639 |
-
|
|
|
640 |
@inproceedings{quan-etal-2020-risawoz,
|
641 |
title = "{R}i{SAWOZ}: A Large-Scale Multi-Domain {W}izard-of-{O}z Dataset with Rich Semantic Annotations for Task-Oriented Dialogue Modeling",
|
642 |
author = "Quan, Jun and
|
@@ -653,5 +694,7 @@ cc-by-4.0: Creative Commons Attribution 4.0 International
|
|
653 |
pages = "930--940",
|
654 |
}
|
655 |
```
|
|
|
656 |
### Contributions
|
657 |
-
|
|
|
|
20 |
- dialogue-modeling
|
21 |
---
|
22 |
# Dataset Card for RiSAWOZ
|
23 |
+
|
24 |
## Table of Contents
|
25 |
+
|
26 |
- [Dataset Card for RiSAWOZ](#dataset-card-for-risawoz)
|
27 |
- [Table of Contents](#table-of-contents)
|
28 |
- [Dataset Description](#dataset-description)
|
|
|
47 |
- [Licensing Information](#licensing-information)
|
48 |
- [Citation Information](#citation-information)
|
49 |
- [Contributions](#contributions)
|
50 |
+
|
51 |
## Dataset Description
|
52 |
+
|
53 |
+
- **Homepage:** <https://terryqj0107.github.io/RiSAWOZ_webpage>
|
54 |
+
- **Repository:** <https://github.com/terryqj0107/RiSAWOZ>
|
55 |
+
- **Paper:** <https://aclanthology.org/2020.emnlp-main.67.pdf>
|
56 |
- **Point of Contact:** Deyi Xiong (dyxiong@tju.edu.cn)
|
57 |
+
|
58 |
### Dataset Summary
|
59 |
+
|
60 |
RiSAWOZ contains 11.2K human-to-human (H2H) multiturn semantically annotated dialogues, with more than 150K utterances spanning over 12 domains, which is larger than all previous annotated H2H conversational datasets. Both single- and multi-domain dialogues are constructed, accounting for 65% and 35%, respectively.
|
61 |
+
|
62 |
### Supported Tasks and Leaderboards
|
63 |
+
|
64 |
- Natural Language Understanding
|
65 |
- Dialogue State Tracking
|
66 |
- Dialogue Context-to-Text Generation
|
67 |
- Coreference Resolution
|
68 |
- Unified Generative Ellipsis and Coreference Resolution
|
69 |
+
|
70 |
### Languages
|
71 |
+
|
72 |
+
Mandarin Chinese (zh-CN)
|
73 |
+
|
74 |
## Dataset Structure
|
75 |
+
|
76 |
### Data Instances
|
77 |
+
|
78 |
A JSON formatted example of a typical instance in RiSAWOZ dataset:
|
79 |
+
|
80 |
+
```JSON
|
81 |
{
|
82 |
"dialogue_id": "attraction_goal_4-63###6177",
|
83 |
"goal": "attraction_goal_4-63: 你是苏州人,但不怎么出去玩。你朋友来苏州找你,你准备带他逛逛“水乡古镇”,你希望客服给你推荐个消费水平“中等”的地方。然后你要问清楚这地方“是否地铁直达”、“特点”、“门票价格”这些信息。最后,你要感谢客服的帮助,然后说再见。",
|
|
|
591 |
]
|
592 |
}
|
593 |
```
|
594 |
+
|
595 |
### Data Fields
|
596 |
+
|
597 |
The data fields are the same among all splits.
|
598 |
+
|
599 |
- dialogue_id (string): dialogue ID
|
600 |
- goal (string): natural language descriptions of the user goal
|
601 |
- domains (list of strings): domains mentioned in current dialogue session
|
602 |
- dialogue (list of dicts): dialog turns and corresponding annotations. Each turn includes:
|
603 |
+
- turn_id (int): turn ID
|
604 |
+
- turn_domain (list of strings): domain mentioned in current turn
|
605 |
+
- user_utterance (string): user utterance
|
606 |
+
- system_utterance (string): system utterance
|
607 |
+
- belief_state (dict): dialogue state, including:
|
608 |
+
- inform slot-values (dict): the slots and corresponding values informed until current turn
|
609 |
+
- turn_inform (dict): the slots and corresponding values informed in current turn
|
610 |
+
- turn request (dict): the slots requested in current turn
|
611 |
+
- user_actions (list of lists): user dialogue acts in current turn
|
612 |
+
- user_actions (list of lists): system dialogue acts in current turn
|
613 |
+
- db_results (list of strings): database search results
|
614 |
+
- segmented_user_utterance (string): word segmentation result of user utterance
|
615 |
+
- segmented_system_utterance (string): word segmentation result of system utterance
|
616 |
+
|
617 |
### Data Splits
|
618 |
+
|
619 |
| | train | validation | test |
|
620 |
| --------- | -----: | ---------: | ----: |
|
621 |
| #dialogues | 10000 | 600 | 600 |
|
622 |
| #turns | 134580 | 8116 | 9286 |
|
623 |
| Avg. turns | 13.5 | 13.5 | 15.5 |
|
624 |
+
|
625 |
## Dataset Creation
|
626 |
+
|
627 |
### Curation Rationale
|
628 |
+
|
629 |
Gather human-to-human dialog in Chinese.
|
630 |
+
|
631 |
### Communicative Goal
|
632 |
+
|
633 |
Generate system response given dialogue context across multiple domains.
|
634 |
+
|
635 |
### Annotations
|
636 |
+
|
637 |
#### Annotation process
|
638 |
+
|
639 |
- Does the dataset have additional annotations for each instance?
|
640 |
+
- crowd-sourced
|
641 |
- What is the number of raters
|
642 |
+
- 51<n<100
|
643 |
- Describe the qualifications required of an annotator.
|
644 |
+
- Chinese native speaker
|
645 |
- How many annotators saw each training example?
|
646 |
+
- 3
|
647 |
- How many annotators saw each test example?
|
648 |
+
- 3
|
649 |
+
|
650 |
### Personal and Sensitive Information
|
651 |
+
|
652 |
The slots and values as well as utterances do not contain any personal information.
|
653 |
+
|
654 |
## Considerations for Using the Data
|
655 |
+
|
656 |
### Social Impact of Dataset
|
657 |
+
|
658 |
RiSAWOZ is the first large-scale multi-domain Chinese Wizard-of-Oz dataset with rich semantic annotations.
|
659 |
+
|
660 |
## Additional Information
|
661 |
+
|
662 |
### Dataset Curators
|
663 |
+
|
664 |
- Jun Quan (Soochow University, Suzhou, China)
|
665 |
- Shian Zhang (Soochow University, Suzhou, China)
|
666 |
- Qian Cao (Soochow University, Suzhou, China)
|
667 |
- Zizhong Li (Tianjin University, Tianjin, China)
|
668 |
- Deyi Xiong (Tianjin University, Tianjin, China)
|
669 |
+
|
670 |
### Funding Information
|
671 |
+
|
672 |
the National Key Research and Development Project
|
673 |
+
|
674 |
### Licensing Information
|
675 |
+
|
676 |
cc-by-4.0: Creative Commons Attribution 4.0 International
|
677 |
+
|
678 |
### Citation Information
|
679 |
+
|
680 |
+
```BibTeX
|
681 |
@inproceedings{quan-etal-2020-risawoz,
|
682 |
title = "{R}i{SAWOZ}: A Large-Scale Multi-Domain {W}izard-of-{O}z Dataset with Rich Semantic Annotations for Task-Oriented Dialogue Modeling",
|
683 |
author = "Quan, Jun and
|
|
|
694 |
pages = "930--940",
|
695 |
}
|
696 |
```
|
697 |
+
|
698 |
### Contributions
|
699 |
+
|
700 |
+
Thanks to Tianhao Shen, Chaobin You and Deyi Xiong for adding this dataset to GEM.
|