Datasets:
Tasks:
Question Answering
Sub-tasks:
extractive-qa
Languages:
English
Multilinguality:
monolingual
Size Categories:
10K<n<100K
Annotations Creators:
crowdsourced
Source Datasets:
extended|sharc
ArXiv:
License:
Commit
•
0ddc52b
1
Parent(s):
4365cab
Fix typo in train split name (#3751)
Browse files* Fix typo in README guide
* Fix split naming and table alignment in README guide
* Fix split naming and table alignment in all datasets
* Fix previously malformed dataset cards
* Fix previously malformed dataset cards
Commit from https://github.com/huggingface/datasets/commit/af6e54febe3b822118dfed28ddb797d8350e5d27
README.md
CHANGED
@@ -23,7 +23,7 @@ paperswithcode_id: null
|
|
23 |
pretty_name: SharcModified
|
24 |
---
|
25 |
|
26 |
-
# Dataset Card
|
27 |
|
28 |
## Table of Contents
|
29 |
- [Dataset Description](#dataset-description)
|
@@ -59,7 +59,14 @@ pretty_name: SharcModified
|
|
59 |
|
60 |
### Dataset Summary
|
61 |
|
62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
### Supported Tasks and Leaderboards
|
65 |
|
@@ -154,9 +161,10 @@ Example of one instance:
|
|
154 |
### Data Splits
|
155 |
|
156 |
The dataset is split into training and validation splits.
|
157 |
-
|
158 |
-
|
|
159 |
-
|
|
|
160 |
|
161 |
## Dataset Creation
|
162 |
|
@@ -214,7 +222,7 @@ The dataset is split into training and validation splits.
|
|
214 |
|
215 |
### Licensing Information
|
216 |
|
217 |
-
|
218 |
|
219 |
### Citation Information
|
220 |
|
23 |
pretty_name: SharcModified
|
24 |
---
|
25 |
|
26 |
+
# Dataset Card for SharcModified
|
27 |
|
28 |
## Table of Contents
|
29 |
- [Dataset Description](#dataset-description)
|
59 |
|
60 |
### Dataset Summary
|
61 |
|
62 |
+
ShARC, a conversational QA task, requires a system to answer user questions based on rules expressed in natural language text.
|
63 |
+
However, it is found that in the ShARC dataset there are multiple spurious patterns that could be exploited by neural models.
|
64 |
+
SharcModified is a new dataset which reduces the patterns identified in the original dataset.
|
65 |
+
To reduce the sensitivity of neural models, for each occurence of an instance conforming to any of the patterns,
|
66 |
+
we automatically construct alternatives where we choose to either replace the current instance with an alternative
|
67 |
+
instance which does not exhibit the pattern; or retain the original instance.
|
68 |
+
The modified ShARC has two versions sharc-mod and history-shuffled.
|
69 |
+
|
70 |
|
71 |
### Supported Tasks and Leaderboards
|
72 |
|
161 |
### Data Splits
|
162 |
|
163 |
The dataset is split into training and validation splits.
|
164 |
+
|
165 |
+
| | train | validation |
|
166 |
+
|--------------|------:|-----------:|
|
167 |
+
| N. Instances | 7638 | 1355 |
|
168 |
|
169 |
## Dataset Creation
|
170 |
|
222 |
|
223 |
### Licensing Information
|
224 |
|
225 |
+
Unknown.
|
226 |
|
227 |
### Citation Information
|
228 |
|