Pretrained model for lemmatization.
Models
We provide a model that enables lemmatization of Swedish text following the SUC3 standard. Note that SUC3 lemmatization does not exactly match the SALDO standard that is used in our Korp resources.
SUC3 was randomly split into training, validation and test sets (80:10:10). The model was trained for 30 epochs using the default Stanza settings. The accuracy on the test set is 99.18.