Commit
•
9ce31b1
1
Parent(s):
4061b15
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,14 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
Task: MCQ with multiple correct answers.
|
| 6 |
+
Dataset: Recommendation of datasets to validate a research question.
|
| 7 |
+
|
| 8 |
+
This dataset is derived from the [DataFinder](https://aclanthology.org/2023.acl-long.573/) dataset. We curate the abstracts of each dataset from [PapersWithCode](https://paperswithcode.com/datasets).
|
| 9 |
+
Given is a short `query` discussing a research question, and keyphrases relevant the query.
|
| 10 |
+
The original training set of the DataFinder dataset has positive and negative candidates for each query, to train a contrastive model.
|
| 11 |
+
We objective is to convert the dataset into a MCQ question-answering task with multiple correct answers. We also add the abstracts from the research papers introducing the datasets so that context can be provided to the models.
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
To reproduce the construction of this dataset, please visit [https://github.com/shruti-singh/scidata_recommendation](https://github.com/shruti-singh/scidata_recommendation).
|