mT5-small based spanish paraphraser Original model Google's mT5 Datasets used for training: spanish PAWS-X Custom database: "Poor-man's" translation of duplicated questions in Quora (translated with Helsinki-NLP/opus-mt-en-es)