---
license: apache-2.0
task_categories:
- text-classification
language:
- ko
dataset_info:
  features:
  - name: comment
    dtype: string
  - name: preference
    dtype: int64
  - name: profanity
    dtype: int64
  - name: gender
    dtype: int64
  - name: politics
    dtype: int64
  - name: nation
    dtype: int64
  - name: race
    dtype: int64
  - name: region
    dtype: int64
  - name: generation
    dtype: int64
  - name: social_hierarchy
    dtype: int64
  - name: appearance
    dtype: int64
  - name: others
    dtype: int64
  splits:
  - name: train
    num_bytes: 7458552
    num_examples: 38361
  - name: test
    num_bytes: 412144
    num_examples: 2000
  download_size: 2947880
  dataset_size: 7870696
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
size_categories:
- 1M<n<10M
---

# KoMultiText: Korean Multi-task Dataset for Classifying Biased Speech

## Dataset Summary

**KoMultiText** is a comprehensive Korean multi-task text dataset designed for classifying biased and harmful speech in online platforms. The dataset focuses on tasks such as **Preference Detection**, **Profanity Identification**, and **Bias Classification** across multiple domains, enabling state-of-the-art language models to perform multi-task learning for socially responsible AI applications.

### Key Features
- **Large-Scale Dataset**: Contains 150,000 comments, including labeled and unlabeled data.
- **Multi-task Annotations**: Covers Preference, Profanity, and nine distinct types of Bias.
- **Human-Labeled**: All labeled data is annotated by **five human experts** to ensure high-quality and unbiased annotations.
- **Real-world Relevance**: Collected from "Real-time Best Gallery" of [DC Inside](https://www.dcinside.com/), a popular online community in South Korea.

### Labels
<img src="https://raw.githubusercontent.com/Dasol-Choi/KoMultiText/main/resources/dataset_configuration.png" width="700">

---
# Dataset Creation

## Source Data
- **Origin**: Comments collected from "Real-time Best Gallery" on [DC Inside](https://www.dcinside.com/).
- **Annotation Process**:
  - **Human Annotation**: Five human annotators independently labeled all comments in the dataset to ensure accuracy and minimize bias.
  - **Labeling Process**: Annotators followed strict guidelines to classify comments into Preference, Profanity, and nine types of Bias. Discrepancies were resolved through majority voting and discussion.
- **Dataset Composition**:
  - **Labeled Data**: 40,361 comments (train/test split).
  - **Unlabeled Data**: 110,000 comments for potential pretraining or unsupervised learning.  [Veiw Dataset](https://huggingface.co/datasets/Dasool/DC_inside_comments)

## How to Load the Dataset

```python
from datasets import load_dataset

# Load the dataset 
dataset = load_dataset("Dasool/KoMultiText")

# Access train and test splits
train_dataset = dataset["train"]
test_dataset = dataset["test"]
```

## Code 
- Korean BERT-based fine-tuning code. [Github](https://github.com/Dasol-Choi/KoMultiText?tab=readme-ov-file)

## Citation

<pre>
@misc{choi2023largescale,
      title={Large-Scale Korean Text Dataset for Classifying Biased Speech in Real-World Online Services}, 
      author={Dasol Choi and Jooyoung Song and Eunsun Lee and Jinwoo Seo and Heejune Park and Dongbin Na},
      year={2023},
      eprint={2310.04313},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
</pre>


## Contact
- dasolchoi@yonsei.ac.kr