classla/FRENK-hate-hr

---
language:
- hr
license:
- other
size_categories:
- 1K<n<10K
task_categories:
- text-classification
task_ids: []
tags:
- hate-speech-detection
- offensive-language
---

# Offensive language dataset of Croatian comments FRENK 1.0

Croatian subset of the [FRENK dataset](http://hdl.handle.net/11356/1433). Also available on HuggingFace dataset hub: [English subset](https://huggingface.co/datasets/5roop/FRENK-hate-en), [Slovenian subset](https://huggingface.co/datasets/5roop/FRENK-hate-sl).

## Dataset Description

- **Homepage:** http://hdl.handle.net/11356/1433
- **Repository:** http://hdl.handle.net/11356/1433
- **Paper:** https://arxiv.org/abs/1906.02045
- **Project page** https://nl.ijs.si/frenk/

## Description of the original dataset

>The original FRENK dataset consists of comments to Facebook posts (news articles) of mainstream media outlets from Croatia, Great Britain, and Slovenia, on the topics of migrants and LGBT. The dataset contains whole discussion threads. Each comment is annotated by the type of socially unacceptable discourse (e.g., inappropriate, offensive, violent speech) and its target (e.g., migrants/LGBT, commenters, media). The annotation schema is described in detail in [https://arxiv.org/pdf/1906.02045.pdf]. Usernames in the metadata are pseudo-anonymised and removed from the comments.
>
>The data in each language (Croatian (hr), English (en), Slovenian (sl), and topic (migrants, LGBT) is divided into a training and a testing portion. The training and testing data consist of separate discussion threads, i.e., there is no cross-discussion-thread contamination between training and testing data. The sizes of the splits are the following: Croatian, migrants: 4356 training comments, 978 testing comments; Croatian LGBT: 4494 training comments, 1142 comments; English, migrants: 4540 training comments, 1285 testing comments; English, LGBT: 4819 training comments, 1017 testing comments; Slovenian, migrants: 5145 training comments, 1277 testing comments; Slovenian, LGBT: 2842 training comments, 900 testing comments.

For this dataset only the Croatian data was used. Training segment has been split into beginning 90% (published here as training split) and end 10% (published here as dev split). Test segment has been preserved in its original form.

## Usage in `Transformers`

```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-hr","binary")
```

For binary classification the following encoding is used:

```python
_CLASS_MAP_BINARY = {
'Acceptable': 0,
'Offensive': 1,
}
```
The original labels are available if the dataset is loaded with the `multiclass` option:

```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-hr","multiclass").
```

In this case the encoding used is:
```python
_CLASS_MAP_MULTICLASS = {
'Acceptable speech': 0,
'Inappropriate': 1,
'Background offensive': 2,
'Other offensive': 3,
'Background violence': 4,
'Other violence': 5,
}
```

## Data structure

* `text`: text
* `target`: who is the target of the hate-speech text ("no target", "commenter", "target" (migrants or LGBT, depending on the topic), or "related to" (again, the topic))
* `topic`: whether the text relates to lgbt or migrants hate-speech domains
* `label`: label of the text instance, see above.

## Data instance

```
{'text': 'Potpisujem komentar g ankice pavicic',
'target': 'No target',
'topic': 'lgbt',
'label': 0}
```

## Licensing information

CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0

## Citation information

When using this dataset please cite the following paper:

```
@misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
}
```

The original dataset can be cited as

```
@misc{11356/1433,
title = {Offensive language dataset of Croatian, English and Slovenian comments {FRENK} 1.0},
author = {Ljube{\v s}i{\'c}, Nikola and Fi{\v s}er, Darja and Erjavec, Toma{\v z}},
url = {http://hdl.handle.net/11356/1433},
note = {Slovenian language resource repository {CLARIN}.{SI}},
copyright = {{CLARIN}.{SI} Licence {ACA} {ID}-{BY}-{NC}-{INF}-{NORED} 1.0},
year = {2021} }
```

原始信息汇总

数据集概述

数据集名称

• 名称： Offensive language dataset of Croatian comments FRENK 1.0

语言

• 语言： Croatian (hr)

许可

• 许可： CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0

大小

• 大小类别： 1K<n<10K

任务类别

• 任务类别： text-classification
• 任务标签： hate-speech-detection, offensive-language

数据集描述

• 原始数据集描述： 包含针对Facebook新闻文章评论的讨论线程，涉及移民和LGBT主题。每条评论根据社会不可接受的话语类型（如不适当、攻击性、暴力言论）及其目标（如移民/LGBT、评论者、媒体）进行标注。
• 数据分割： 训练数据分为前90%（训练分割）和后10%（开发分割），测试数据保持原样。

数据结构

• 字段：
  • text: 文本内容
  • target: 仇恨言论的目标（无目标、评论者、目标（移民或LGBT）、相关（主题））
  • topic: 文本关联的仇恨言论领域（LGBT或移民）
  • label: 文本实例的标签

使用示例

• 二元分类编码：
  • Acceptable: 0
  • Offensive: 1
• 多类别分类编码：
  • Acceptable speech: 0
  • Inappropriate: 1
  • Background offensive: 2
  • Other offensive: 3
  • Background violence: 4
  • Other violence: 5

引用信息

• 引用论文：

  @misc{ljubešić2019frenk, title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English}, author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec}, year={2019}, eprint={1906.02045}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/1906.02045} }

• 原始数据集引用：

  @misc{11356/1433, title={Offensive language dataset of Croatian, English and Slovenian comments FRENK 1.0}, author={Ljube{v s}i{c}, Nikola and Fi{v s}er, Darja and Erjavec, Toma{v z}}, url={http://hdl.handle.net/11356/1433}, note={Slovenian language resource repository CLARIN.SI}, copyright={CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0}, year={2021} }