---
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