Datasets:
Languages:
Indonesian
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,19 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-sa-4.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-sa-4.0
|
| 3 |
---
|
| 4 |
+
# IndoCloze
|
| 5 |
+
|
| 6 |
+
## About
|
| 7 |
+
We hired seven Indonesian university students to each write 500 short stories over a period of one month. This paper wins **Best Paper Award at CSRR (ACL 2022)**.
|
| 8 |
+
|
| 9 |
+
## Paper
|
| 10 |
+
Fajri Koto, Timothy Baldwin, and Jey Han Lau. [_Cloze Evaluation for Deeper Understanding of Commonsense Stories in Indonesian_](https://aclanthology.org/2022.csrr-1.2.pdf).
|
| 11 |
+
In In Proceedings of Commonsense Representation and Reasoning Workshop 2022 (**CSRR at ACL 2022**), Dublin, Ireland.
|
| 12 |
+
|
| 13 |
+
## Dataset
|
| 14 |
+
|
| 15 |
+
A story in our dataset consists of four-sentence premise, one-sentence correct ending, and one-sentence incorrect ending. In total, we have created 2,325 Indonesian stories with the train/dev/test split 1,000/200/1,135. Please see some examples of our data below, and note that the English translation is only for the illustratrive purposes.
|
| 16 |
+
|
| 17 |
+
<h3 align="center">
|
| 18 |
+
<img src="https://raw.githubusercontent.com/fajri91/eval_picts/master/indocloze.png" width="850">
|
| 19 |
+
</h3>
|