README.md

metadata

license: apache-2.0
task_categories:
  - text-classification
language:
  - en

• 36.528 English texts in total, 12.955 NOT offensive and 23.573O OFFENSIVE texts
• All duplicate values were removed
• Split using sklearn into 80% train and 20% temporary test (stratified label). Then split the test set using 0.50% test and validation (stratified label)
• Split: 80/10/10
• Train set label distribution: 0 ==> 10.364, 1 ==> 18.858
• Validation set label distribution: 0 ==> 1.296, 1 ==> 2.357
• Test set label distribution: 0 ==> 1.295, 1 ==> 2.358
• The OLID dataset (Zampieri et al., 2019) and the labels "Offensive" and "Neither" from the paper's dataset "Automated Hate Speech Detection and the Problem of Offensive Language" (Davidson et al.,2017)