---
language:
- tr
license:
- cc-by-2.0
---

# OffensEval-TR 2020

This repository hosts a special version of the well-known OffensEval-TR 2020 dataset.

This version splits the official training dataset into a slightly smaller training dataset and a development set. Purpose of this special version is to have a more consistent dataset for benchmarks.

## Original Dataset Description

- **Homepage:** [offensive-turkish](https://coltekin.github.io/offensive-turkish/)
- **Paper:** [A Corpus of Turkish Offensive Language on Social Media](https://coltekin.github.io/offensive-turkish/troff.pdf)
- **Point of Contact:** [Çağrı Çöltekin](ccoltekin@sfs.uni-tuebingen.de)
- **Original Dataset on Hugging Face Model Hub:** https://huggingface.co/datasets/coltekin/offenseval2020_tr

### Dataset Summary

The original training dataset contains 31,756 annotated tweets. Our new training dataset has 30,000 annotated tweets, whereas the development dataset has 1,756 tweets.

The test dataset remains unmodified and contains 3,528 tweets.

Twitter user mentions were substituted by `@USER` and URLs have been substitute by `URL`.

Each instance contains up to 1 labels corresponding to one of the following sub-task:

- Sub-task A: Offensive language identification; 

### Supported Tasks and Leaderboards

The dataset was published on this [paper](https://coltekin.github.io/offensive-turkish/troff.pdf). 

### Languages

The dataset is based on Turkish.

## Dataset Structure

### Data Instances

A binary dataset with with (`NOT`) Not Offensive and (`OFF`) Offensive tweets.  

### Loader

An example of how to use this dataset within the awesome Flair library can be found in [this](FlairDatasetLoader.ipynb) notebook.

# Acknowledgements

Many thanks to [Cagri Cöltekin](https://huggingface.co/coltekin) for introducing this great dataset and also releasing it on the Model Hub. Model card is slightly adopted and modified from Cagri's [model card](https://huggingface.co/datasets/coltekin/offenseval2020_tr).