Polish-Kashubian parallel translation corpus

The dataset contains Polish words and sentences and their translations into Kashubian. The dataset consists of train and test subsets. The train subset contains about 100,000 parallel translations. It was created using two types of sources. The first one is the online dictionaries: kaszebe.org sloworz.org odmiana.net The second type of source was an existing dataset that was incorporated into this one: OPUS Tatoeba Challenge The dataset was carefully cleaned and duplicates were removed.The test dataset is distributed together with the training dataset. It contains 70 parallel Poish-Kashubian sentences. The sentences were obtained from sources other than the train data to make them reliable.