ArneBinder
commited on
Commit
•
b692ce8
1
Parent(s):
d97f676
add images with examples
Browse files- README.md +43 -11
- img/as19f1.png +3 -0
- img/scidtb-sam.png +3 -0
README.md
CHANGED
@@ -2,9 +2,8 @@
|
|
2 |
|
3 |
### Dataset Summary
|
4 |
|
5 |
-
Built from 60 English scientific abstracts in the larger annotated dataset
|
6 |
-
|
7 |
-
The dataset contains 327 sentences, 8012 tokens, 862 discourse units and 352 argumentative units (of 6 labels) linked by 292 argumentative relations (of 5 labels).
|
8 |
|
9 |
### Supported Tasks and Leaderboards
|
10 |
|
@@ -13,13 +12,13 @@ The dataset contains 327 sentences, 8012 tokens, 862 discourse units and 352 arg
|
|
13 |
|
14 |
### Languages
|
15 |
|
16 |
-
The language in the dataset is English (academic).
|
17 |
|
18 |
## Dataset Structure
|
19 |
|
20 |
### Data Instances
|
21 |
|
22 |
-
- **Size of downloaded dataset files:** 32.4 KB
|
23 |
|
24 |
```
|
25 |
{
|
@@ -46,11 +45,44 @@ The language in the dataset is English (academic).
|
|
46 |
|
47 |
### Data Splits
|
48 |
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
## Dataset Creation
|
56 |
|
@@ -62,7 +94,7 @@ The language in the dataset is English (academic).
|
|
62 |
|
63 |
### Source Data
|
64 |
|
65 |
-
The source data is available online at https://emnlp2014.org/.
|
66 |
|
67 |
#### Initial Data Collection and Normalization
|
68 |
|
|
|
2 |
|
3 |
### Dataset Summary
|
4 |
|
5 |
+
Built from 60 English scientific abstracts in the larger annotated dataset: *Discourse Dependency TreeBank for Scientific Abstracts* (SciDTB; [Yang & Li, 2018](https://aclanthology.org/P18-2071)), [Accuosto and Saggion (2019)](https://aclanthology.org/W19-4505.pdf) offered a fine-grained annotated dataset for the argumentative component classification and relation classification tasks, based on the study of Kirschner et al. ([2015](https://aclanthology.org/W15-0501/)).
|
6 |
+
The dataset is token-based and segmented into elementary units that cover that entire text.
|
|
|
7 |
|
8 |
### Supported Tasks and Leaderboards
|
9 |
|
|
|
12 |
|
13 |
### Languages
|
14 |
|
15 |
+
The language in the dataset is English (academic; computational linguistics).
|
16 |
|
17 |
## Dataset Structure
|
18 |
|
19 |
### Data Instances
|
20 |
|
21 |
+
- **Size of downloaded dataset files:** 32.4 KB ([link](http://scientmin.taln.upf.edu/argmin/scidtb_argmin_annotations.tgz))
|
22 |
|
23 |
```
|
24 |
{
|
|
|
45 |
|
46 |
### Data Splits
|
47 |
|
48 |
+
This dataset contains 60 documents, and it is not pre-split.
|
49 |
+
|
50 |
+
### Label Description and Statistics
|
51 |
+
|
52 |
+
#### Components
|
53 |
+
|
54 |
+
| | train | percentage |
|
55 |
+
| -------------------------------------------------------------------------------------- | ------------------------------------: | ----------------------------------------------------------: |
|
56 |
+
| `proposal`<br/>`mean`<br/>`result`<br/>`observation`<br/>`assertion`<br/>`description` | 110<br/>63<br/>74<br/>11<br/>88<br/>7 | 31.2 %<br/>17.8 %<br/>21.0 %<br/>3.1 %<br/>24.9 %<br/>2.0 % |
|
57 |
+
|
58 |
+
- `proposal`: problem or approach
|
59 |
+
- `assertion`: conclusion or known fact
|
60 |
+
- `result`: interpretation of data
|
61 |
+
- `observation`: data
|
62 |
+
- `means`: implementation
|
63 |
+
- `description`: definitions/other information
|
64 |
+
|
65 |
+
(Accuosto & Saggion, 2019, p. 44)
|
66 |
+
|
67 |
+
#### Relations
|
68 |
+
|
69 |
+
| | train | percentage |
|
70 |
+
| ------------------------------------------------------------------- | ------------------------------: | --------------------------------------------: |
|
71 |
+
| `support`<br/>`attack`<br/>`detail`<br/>`sequence`<br/>`additional` | 126<br/>0<br/>129<br/>11<br/>27 | 43.0 %<br/>0 %<br/>22.0 %<br/>3.8 %<br/>9.2 % |
|
72 |
+
|
73 |
+
- `support`: between an argument component A and another argument component B indicates that A supports (reasons, proves) B.
|
74 |
+
- `attack`: if A attacks (restricts, contradicts) B. However, the authors claimed that they did not identify any attack in the current texts (2020, p.44).
|
75 |
+
- `additional`: *not explicitly stated*, but possibly refers to discourse relation in Yang & Li (2018, p.445), in other words, elaboration
|
76 |
+
- `detail`: if A is a detail of B and gives more information or defines something stated in B without argumentative reason
|
77 |
+
- `sequence`: if two (or more) argument components belong together and only make sense in combination, i.e., they form a multi-sentence argument component.
|
78 |
+
|
79 |
+
(Kirschner et al., 2015, p. 5)
|
80 |
+
|
81 |
+
#### Examples
|
82 |
+
|
83 |
+
![sample1](img/as19f1.png)
|
84 |
+
(Accuosto & Saggion, 2019, p.44, Figure 1)
|
85 |
+
![sample2](img/scidtb-sam.png)
|
86 |
|
87 |
## Dataset Creation
|
88 |
|
|
|
94 |
|
95 |
### Source Data
|
96 |
|
97 |
+
The source data is a subset of EMNLP abstracts in the year 2014, available online at https://emnlp2014.org/.
|
98 |
|
99 |
#### Initial Data Collection and Normalization
|
100 |
|
img/as19f1.png
ADDED
Git LFS Details
|
img/scidtb-sam.png
ADDED
Git LFS Details
|