dimuthulk/sinhala-offensive-small-dataset

📊 Dataset Card for Sinhala Offensive Small Dataset

📝 Dataset Description

• —Repository: dimuthulk/sinhala-offensive-small-dataset
• —Language(s) (NLP): Sinhala (si)
• —License: Apache 2.0

📋 Dataset Summary

This dataset contains Sinhala text data categorized for offensive language detection. It was developed as the primary classification dataset for the first phase of the "Sinhala Offensive Text Detoxification Pipeline" research project conducted at the University of Kelaniya.

The dataset is intended to help researchers and developers train Natural Language Processing (NLP) models to identify offensive, toxic, or abusive content in the Sinhala language.

🚀 Uses

✅ Direct Use

This dataset is suitable for training and evaluating text classification models (e.g., BERT, RoBERTa, or custom architectures) to detect offensive language in Sinhala text.

❌ Out-of-Scope Use

This dataset should not be used to train generative models to produce offensive content or to target specific individuals/groups.

🗂️ Dataset Structure

🔍 Data Instances

A typical instance in the dataset consists of a text snippet in Sinhala and its corresponding label indicating whether the text is offensive or not.

🏷️ Data Fields

• —text (string): A string containing the Sinhala text to be classified.
• —label (string): A string representing the classification (e.g., OFF for offensive language).

🛠️ Dataset Creation

📂 Source Data

The data was collected and curated to reflect natural Sinhala language usage, specifically focusing on identifying abusive or offensive patterns in digital communication.

ℹ️ Additional Information

🎓 Dataset Curators

Dimuthu Rathnayaka (Undergraduate, Electronics and Computer Science, University of Kelaniya)

📚 Citation Information

If you use this dataset in your research, please cite it as follows:

@misc{sinhala_offensive_small_dataset_2026,
  author = {Rathnayaka, Dimuthu},
  title = {Sinhala Offensive Small Dataset},
  year = {2026},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  howpublished = {\url{[https://huggingface.co/datasets/dimuthulk/sinhala-offensive-small-dataset](https://huggingface.co/datasets/dimuthulk/sinhala-offensive-small-dataset)}}
}