Datasets:

benoitfavre
/

Abuseval

dataset_info:
  features:
    - name: text
      dtype: string
    - name: label
      dtype: string
  splits:
    - name: train
      num_bytes: 1583633
      num_examples: 11240
    - name: valid
      num_bytes: 144044
      num_examples: 1000
    - name: test
      num_bytes: 141548
      num_examples: 1000
  download_size: 1156986
  dataset_size: 1869225
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: valid
        path: data/valid-*
      - split: test
        path: data/test-*

Abuseval

Note: this is not the official valid/test split.

Abusive language detection, with three labels: NOTABU (not abusive), EXP (explicit), IMP (implicit)

English tweets. User names, urls, etc have been replaced. See: https://github.com/tommasoc80/AbuseEval, https://github.com/joeykay9/offenseval