Bert-base-german-cased model fine-tuned on a corpus of German-Swiss.
Bert-base-german-cased model fine-tuned on a corpus of German-Swiss.
Fine-tuning was done on the Swiss German data of the
Leipzig Corpora Collection and
SwissCrawl. For testing purposes, the model was evaluated on the
Swiss Dialect Classification dataset as down-stream task.
It outperformed its parent model (
bert-base-german-cased) by approx. 5% accuracy.
