Full model card will be added soon.
"german-nlp-group/electra-base-german-uncased"-Model finetuned on an offensive language task. The dataset of 1250 samples was manually labeled into two categories, namely "HOF" (Hate or Offensive Language) and "NOT2 (No Hate or Offensive Language). Anntation guidelines were taken from the Germeval 2019 Task 2 (https://www.zora.uzh.ch/id/eprint/178687/1/GermEvalSharedTask2019Iggsa.pdf). The model achieved an F1-Score of 0.89 on the test data set.
A detailed model card will be added soon with a more precise evaluation and with the train/test datasets.
- Downloads last month