---
license: mit
language:
- kk
task_categories:
- text-classification
pretty_name: Kazakh Swear Words Dataset
size_categories:
- n<1K
tags:
- profanity
- content-moderation
- kazakh
- low-resource
- toxicity-detection
dataset_info:
  features:
  - name: text
    dtype: string
  - name: label
    dtype: int64
  - name: category
    dtype: string
  - name: language
    dtype: string
  - name: severity
    dtype: string
---

# Kazakh Swear Words Dataset 🇰🇿

Dataset of Kazakh obscene and profane expressions for NLP tasks including text classification, content moderation, toxicity detection, and LLM fine-tuning.

## Dataset Description

This is a low-resource language dataset containing Kazakh profanity, swear words, and offensive expressions along with neutral examples for binary classification tasks.

### Languages
- Kazakh (`kk`)

### Dataset Structure

#### Data Files
- **data.jsonl** - Individual words (56 entries)
- **sentences.jsonl** - Phrases and expressions (63 entries)

#### Data Fields

- `text` (string): The Kazakh word or phrase
- `label` (int): Binary classification label
  - `1` - Profanity/Offensive
  - `0` - Neutral
- `category` (string): Type of content
  - `vulgar` - Vulgar/obscene terms
  - `offensive` - Offensive expressions
  - `insult` - Insults
  - `neutral` - Neutral content
- `language` (string): ISO 639-1 language code (`kk`)
- `severity` (string): Severity level
  - `severe` - Highly offensive
  - `moderate` - Moderately offensive
  - `mild` - Mildly offensive
  - `none` - Not offensive

### Dataset Statistics

**data.jsonl (Words)**
- Total: 56 entries
- Profanity: 28 (50%)
- Neutral: 28 (50%)

**sentences.jsonl (Phrases)**
- Total: 63 entries
- Profanity: 38 (60%)
- Neutral: 25 (40%)

### Use Cases
- Content moderation systems
- Chatbot filtering
- Toxicity detection models
- Text classification
- Low-resource language NLP research

### Usage Example

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("json", data_files="data.jsonl")

# Or for sentences
sentences = load_dataset("json", data_files="sentences.jsonl")

# Access data
for item in dataset['train']:
    print(f"Text: {item['text']}, Label: {item['label']}, Category: {item['category']}")
```

### Ethical Considerations

This dataset contains offensive and profane language. It is intended solely for:
- Academic research
- Content moderation system development
- Toxicity detection model training
- NLP research in low-resource languages

**Do not use this dataset to:**
- Generate offensive content
- Harass or harm individuals
- Train models for malicious purposes

### Disclaimer

This dataset contains explicit profanity and offensive language in Kazakh. It is provided for research and content moderation purposes only. Users are responsible for ensuring ethical use of this data.

### License

MIT License

### Citation

If you use this dataset, please cite:

```
@misc{kazakh_swear_words_2025,
  title={Kazakh Swear Words Dataset},
  author={Rustem Kaimolla},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/Rustem-Kaimolla/kazakh-swear-words}}
}
```