Datasets:
Commit
•
041a62d
1
Parent(s):
eb70321
Update README.md
Browse files
README.md
CHANGED
@@ -32,7 +32,88 @@ dataset_info:
|
|
32 |
num_examples: 5571
|
33 |
download_size: 231554886
|
34 |
dataset_size: 1605837137
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
---
|
36 |
-
# Dataset Card for
|
37 |
|
38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
num_examples: 5571
|
33 |
download_size: 231554886
|
34 |
dataset_size: 1605837137
|
35 |
+
license: cc-by-sa-4.0
|
36 |
+
task_categories:
|
37 |
+
- question-answering
|
38 |
+
- conversational
|
39 |
+
- text2text-generation
|
40 |
+
language:
|
41 |
+
- en
|
42 |
+
pretty_name: Canard Wikipedia-augmented
|
43 |
+
size_categories:
|
44 |
+
- 10K<n<100K
|
45 |
---
|
46 |
+
# Dataset Card for Canard_Wiki-augmented
|
47 |
|
48 |
+
### Summary
|
49 |
+
|
50 |
+
This is a dataset of fact-retrieving conversations about Wikipedia articles, with all responses grounded in a specific segment of text in the referenced Wikipedia article.
|
51 |
+
It is an extended version of [Canard](https://sites.google.com/view/qanta/projects/canard)
|
52 |
+
and [QuAC](https://huggingface.co/datasets/quac) datasets,
|
53 |
+
augmented with the contexts of [English Wikipedia](https://huggingface.co/datasets/wikipedia).
|
54 |
+
|
55 |
+
### Supported Tasks
|
56 |
+
|
57 |
+
The dataset is intended to train a factually-consistent conversational model able to ground all its responses to the corresponding source(s).
|
58 |
+
However, the data can also be used to evaluate the information retrieval (IR) system for given queries, for contextual disambiguation of the queries from a conversation, etc.
|
59 |
+
|
60 |
+
## Dataset Structure
|
61 |
+
|
62 |
+
The dataset can be loaded by simply choosing a split (`train` or `test`) and calling:
|
63 |
+
|
64 |
+
```python
|
65 |
+
import datasets
|
66 |
+
canard_augm_test = datasets.load_dataset("gaussalgo/Canard_Wiki-augmented", split="test")
|
67 |
+
|
68 |
+
print(canard_augm_test[0]) # print the first sample
|
69 |
+
```
|
70 |
+
|
71 |
+
### Data Instances
|
72 |
+
|
73 |
+
The samples of Canard_Wiki-augmented have this format:
|
74 |
+
|
75 |
+
```python
|
76 |
+
{'History': ['Anna Politkovskaya', 'The murder remains unsolved, 2016'],
|
77 |
+
'QuAC_dialog_id': 'C_0aaa843df0bd467b96e5a496fc0b033d_1',
|
78 |
+
'Question': 'Did they have any clues?',
|
79 |
+
'Question_no': 1,
|
80 |
+
'answer': 'Her colleagues at Novaya gazeta protested that until the instigator or sponsor of the crime was identified, arrested and prosecuted the case was not closed.'
|
81 |
+
'Rewrite': 'Did investigators have any clues in the unresolved murder of Anna Politkovskaya?',
|
82 |
+
'true_page_title': 'Anna Politkovskaya',
|
83 |
+
'true_contexts': 'In September 2016 Vladimir Markin, official spokesman for (...)',
|
84 |
+
'true_contexts_wiki': 'Anna Stepanovna Politkovskaya was a US-born Russian journalist (...)',
|
85 |
+
'extractive': True
|
86 |
+
'retrieved_contexts': ['Clues was an indie rock band from Montreal, Canada formed by Alden Penner (...)',
|
87 |
+
'High Stakes is a British game show series hosted by Jeremy Kyle, in which (...)']
|
88 |
+
```
|
89 |
+
|
90 |
+
### Data Fields
|
91 |
+
|
92 |
+
* **History**: History of the conversation from Canard. The first two entries of the conversation are always synthetic.
|
93 |
+
* **QuAC_dialog_id**: Dialogue ID mapping the conversation to the original QuAC dataset (*dialogue_id* in QuAC).
|
94 |
+
* **Question**: Current question of the user from Canard.
|
95 |
+
* **Question_no**: Ordering of the user's question from the conversation, originally from Canard.
|
96 |
+
* **answer**: Correctly extracted answer to a given question from a relevant Wikipedia article (*true_contexts*). Note that some of the questions are open, thus the listed answer is not the only correct possibility.
|
97 |
+
* **Rewrite**: A rephrased version of *Question*, manually disambiguated from the context of *History* by the annotators of Canard.
|
98 |
+
* **true_page_title**: Title of the Wikipedia article containing *answer*. *wikipedia_page_title* from QuAC.
|
99 |
+
* **true_contexts**: An excerpt of the paragraph with an answer from the Wikipedia article titled *true_page_title*.
|
100 |
+
* **true_contexts_wiki**: A full contents of Wikipedia article (*text* from Wikipedia dataset), where *true_page_title* matches Wikipedia *title*. Note that the Wikipedia dataset was retrieved on 2nd of April, 2023.
|
101 |
+
* **extractive**: A flag whether the *answer* in this sample can be found as an exact-match in *true_contexts_wiki*.
|
102 |
+
* **retrieved_contexts**: "Distractor" contexts retrieved from the full Wikipedia dataset using the okapi-BM25 IR system on a **Rewrite** question.
|
103 |
+
|
104 |
+
### Data Splits
|
105 |
+
|
106 |
+
* **train** split is aligned with the training splits of Canard and QuAC.
|
107 |
+
* **test** split matches the validation split of QuAC and the test split of Canard (where the conversation ids match).
|
108 |
+
|
109 |
+
## Licensing
|
110 |
+
|
111 |
+
This dataset is composed of [QuAC](https://huggingface.co/datasets/quac) (MIT),
|
112 |
+
[Canard](https://sites.google.com/view/qanta/projects/canard) (CC BY-SA 4.0)
|
113 |
+
and [Wikipedia](https://huggingface.co/datasets/wikipedia) (CC BY SA 3.0).
|
114 |
+
Canard_Wiki-augmented is therefore licensed under CC BY-SA 4.0 as well, allowing it to be also commercially used.
|
115 |
+
|
116 |
+
## Cite
|
117 |
+
|
118 |
+
If you use this dataset in a research, do not forget to cite the authors of original datasets, that Canard_Wiki-augmented is derived from:
|
119 |
+
[QuAC](https://huggingface.co/datasets/quac), [Canard](https://sites.google.com/view/qanta/projects/canard).
|