---
language:
- en
- uk
- ru
- de
- zh
- am
- ar
- hi
- es
license: openrail++
size_categories:
- 1K<n<10K
task_categories:
- text-generation
dataset_info:
  features:
  - name: toxic_sentence
    dtype: string
  - name: neutral_sentence
    dtype: string
  splits:
  - name: en
    num_bytes: 47435
    num_examples: 400
  - name: ru
    num_bytes: 89453
    num_examples: 400
  - name: uk
    num_bytes: 78106
    num_examples: 400
  - name: de
    num_bytes: 86818
    num_examples: 400
  - name: es
    num_bytes: 56868
    num_examples: 400
  - name: am
    num_bytes: 133489
    num_examples: 400
  - name: zh
    num_bytes: 79089
    num_examples: 400
  - name: ar
    num_bytes: 85237
    num_examples: 400
  - name: hi
    num_bytes: 107518
    num_examples: 400
  download_size: 489288
  dataset_size: 764013
configs:
- config_name: default
  data_files:
  - split: en
    path: data/en-*
  - split: ru
    path: data/ru-*
  - split: uk
    path: data/uk-*
  - split: de
    path: data/de-*
  - split: es
    path: data/es-*
  - split: am
    path: data/am-*
  - split: zh
    path: data/zh-*
  - split: ar
    path: data/ar-*
  - split: hi
    path: data/hi-*
---
**MultiParaDetox**

This is the multilingual parallel dataset for text detoxification prepared for [CLEF TextDetox 2024](https://pan.webis.de/clef24/pan24-web/text-detoxification.html) shared task.
For each of 9 languages, we collected 1k pairs of toxic<->detoxified instances splitted into two parts: dev (400 pairs) and test (600 pairs). 

**!!!April, 23rd, update: We are realsing the parallel dev set! The test part for the final phase of the competition is available [here](https://huggingface.co/datasets/textdetox/multilingual_paradetox_test)!!!**

The list of the sources for the original toxic sentences:
* English: [Jigsaw](https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge), [Unitary AI Toxicity Dataset](https://github.com/unitaryai/detoxify)
* Russian: [Russian Language Toxic Comments](https://www.kaggle.com/datasets/blackmoon/russian-language-toxic-comments), [Toxic Russian Comments](https://www.kaggle.com/datasets/alexandersemiletov/toxic-russian-comments)
* Ukrainian: [Ukrainian Twitter texts](https://github.com/saganoren/ukr-twi-corpus)
* Spanish: [Detecting and Monitoring Hate Speech in Twitter](https://www.mdpi.com/1424-8220/19/21/4654), [Detoxis](https://rdcu.be/dwhxH), [RoBERTuito: a pre-trained language model for social media text in Spanish](https://aclanthology.org/2022.lrec-1.785/)
* German: [GemEval 2018, 2021](https://aclanthology.org/2021.germeval-1.1/)
* Amhairc: [Amharic Hate Speech](https://github.com/uhh-lt/AmharicHateSpeech)
* Arabic: [OSACT4](https://edinburghnlp.inf.ed.ac.uk/workshops/OSACT4/)
* Hindi: [Hostility Detection Dataset in Hindi](https://competitions.codalab.org/competitions/26654#learn_the_details-dataset), [Overview of the HASOC track at FIRE 2019: Hate Speech and Offensive Content Identification in Indo-European Languages](https://dl.acm.org/doi/pdf/10.1145/3368567.3368584?download=true)