create readme
Browse files
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
languages:
|
| 3 |
+
- tr
|
| 4 |
+
size_categories:
|
| 5 |
+
- 100K<n<1M
|
| 6 |
+
source_datasets:
|
| 7 |
+
- nli_tr
|
| 8 |
+
task_categories:
|
| 9 |
+
- text-scoring
|
| 10 |
+
task_ids:
|
| 11 |
+
- semantic-similarity-scoring
|
| 12 |
+
---
|
| 13 |
+
# NLI-TR for Supervised SimCSE
|
| 14 |
+
|
| 15 |
+
This dataset is a modified version of [NLI-TR](https://huggingface.co/datasets/nli_tr) dataset. Its intended use is to train Supervised [SimCSE](https://github.com/princeton-nlp/SimCSE) models for sentence-embeddings. Steps followed to produce this dataset are listed below:
|
| 16 |
+
1. Merge train split of snli_tr and multinli_tr subsets.
|
| 17 |
+
2. Find every premise that has an entailment hypothesis **and** a contradiction hypothesis.
|
| 18 |
+
3. Write found triplets into sent0 (premise), sent1 (entailment hypothesis), hard_neg (contradiction hypothesis) format.
|