Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Languages:
English
Size:
10K - 100K
License:
Commit
·
b14e1a9
1
Parent(s):
8697970
Update README.md
Browse files
README.md
CHANGED
@@ -4,6 +4,24 @@ task_categories:
|
|
4 |
- text-classification
|
5 |
language:
|
6 |
- en
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
configs:
|
8 |
- config_name: snli_extended
|
9 |
data_files:
|
@@ -21,19 +39,4 @@ configs:
|
|
21 |
data_files:
|
22 |
- split: test
|
23 |
path: fever.jsonl
|
24 |
-
|
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.
|
|
|
4 |
- text-classification
|
5 |
language:
|
6 |
- en
|
7 |
+
tags:
|
8 |
+
- 'rationale-extraction'
|
9 |
+
- reasoning
|
10 |
+
- nli
|
11 |
+
- fact-checking
|
12 |
+
- explainability
|
13 |
+
pretty_name: spanex
|
14 |
+
size_categories:
|
15 |
+
- 1K<n<10K
|
16 |
+
---
|
17 |
+
|
18 |
+
SpanEx consists of 7071 instances annotated for span interactions.
|
19 |
+
SpanEx is the first dataset with human phrase-level interaction explanations with explicit labels for interaction types.
|
20 |
+
Moreover, SpanEx is annotated by three annotators, which opens new avenues for studies of human explanation agreement -- an understudied area in the explainability literature.
|
21 |
+
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.
|
22 |
+
We collect explanations of span interactions for NLI on the SNLI dataset and for FC on the FEVER dataset.
|
23 |
+
|
24 |
+
---
|
25 |
configs:
|
26 |
- config_name: snli_extended
|
27 |
data_files:
|
|
|
39 |
data_files:
|
40 |
- split: test
|
41 |
path: fever.jsonl
|
42 |
+
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|