Datasets:

sentence-transformers
/

all-nli

Tasks:

Feature Extraction

Sentence Similarity

Modalities:

Formats:

Languages:

Size:

Tags:

sentence-transformers

Libraries:

Dataset card Viewer Files Files and versions Community

tomaarsen HF staff commited on Apr 25

Commit

0c94fa2

•

1 Parent(s): 54d241c

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -176,5 +176,5 @@ Despite originally being intended for Natural Language Inference (NLI), this dat
 ```python
 {'anchor': 'A person on a horse jumps over a broken down airplane.', 'positive': 'A person is outdoors, on a horse.', 'negative': 'A person is at a diner, ordering an omelette.'}
 ```
-* Collection strategy: Reading the SNLI & MultiNLI datasets, for each "premise" making a list of entailing and contradictory sentences using the dataset labels. Then, considering all possible triplets out of these entailing and contradictory lists. The reverse ("entailment" as "anchor" and "premise" as "positive") is also included.
 * Deduplified: Yes

 ```python
 {'anchor': 'A person on a horse jumps over a broken down airplane.', 'positive': 'A person is outdoors, on a horse.', 'negative': 'A person is at a diner, ordering an omelette.'}
 ```
+* Collection strategy: Reading the SNLI & MultiNLI datasets, for each "premise" making a list of entailing and contradictory sentences using the dataset labels. Then, considering all possible triplets out of these entailing and contradictory lists. The reverse ("entailment" as "anchor" and "premise" as "positive") is not included.
 * Deduplified: Yes