skupina-7/nlp-paraphrases-20k-pruned

The dataset contains 20703 records. The dataset was created by removing all dataset items from the original 27k dataset that had a BLEU score 0 or more than 0.3388.