Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Languages:
English
Size:
10K - 100K
License:
Commit
·
8697970
1
Parent(s):
689910f
Update README.md
Browse files
README.md
CHANGED
@@ -1,10 +1,9 @@
|
|
1 |
-
SpanEx consists of 7071 instances annotated for span interactions.
|
2 |
-
SpanEx is the first dataset with human phrase-level interaction explanations with explicit labels for interaction types.
|
3 |
-
Moreover, SpanEx is annotated by three annotators, which opens new avenues for studies of human explanation agreement -- an understudied area in the explainability literature.
|
4 |
-
Our study reveals that while human annotators often agree on span interactions, they also offer complementary reasons for a prediction, collectively providing a comprehensive set of reasons for a prediction.
|
5 |
-
We collect explanations of span interactions for NLI on the SNLI dataset and for FC on the FEVER dataset.
|
6 |
---
|
7 |
license: mit
|
|
|
|
|
|
|
|
|
8 |
configs:
|
9 |
- config_name: snli_extended
|
10 |
data_files:
|
@@ -22,15 +21,19 @@ configs:
|
|
22 |
data_files:
|
23 |
- split: test
|
24 |
path: fever.jsonl
|
25 |
-
task_categories:
|
26 |
-
- text-classification
|
27 |
-
language:
|
28 |
-
- en
|
29 |
tags:
|
30 |
-
-
|
|
|
31 |
- nli
|
|
|
32 |
- explainability
|
33 |
-
pretty_name:
|
34 |
size_categories:
|
35 |
- 1K<n<10K
|
36 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
+
task_categories:
|
4 |
+
- text-classification
|
5 |
+
language:
|
6 |
+
- en
|
7 |
configs:
|
8 |
- config_name: snli_extended
|
9 |
data_files:
|
|
|
21 |
data_files:
|
22 |
- split: test
|
23 |
path: fever.jsonl
|
|
|
|
|
|
|
|
|
24 |
tags:
|
25 |
+
- 'rationale-extraction '
|
26 |
+
- reasoning
|
27 |
- nli
|
28 |
+
- fact-checking
|
29 |
- explainability
|
30 |
+
pretty_name: spanex
|
31 |
size_categories:
|
32 |
- 1K<n<10K
|
33 |
+
---
|
34 |
+
|
35 |
+
SpanEx consists of 7071 instances annotated for span interactions.
|
36 |
+
SpanEx is the first dataset with human phrase-level interaction explanations with explicit labels for interaction types.
|
37 |
+
Moreover, SpanEx is annotated by three annotators, which opens new avenues for studies of human explanation agreement -- an understudied area in the explainability literature.
|
38 |
+
Our study reveals that while human annotators often agree on span interactions, they also offer complementary reasons for a prediction, collectively providing a comprehensive set of reasons for a prediction.
|
39 |
+
We collect explanations of span interactions for NLI on the SNLI dataset and for FC on the FEVER dataset.
|