Azri-Muhsin/sinhala-news-comment-sentiment

---
dataset_info:
features:
- name: article_title
dtype: string
- name: article_author
dtype: string
- name: article_date
dtype: string
- name: article_body
dtype: string
- name: text
dtype: string
- name: label
dtype:
class_label:
names:
'0': negative
'1': neutral
'2': positive
splits:
- name: train
num_bytes: 35141453.94672856
num_examples: 7238
- name: test
num_bytes: 8787791.05327144
num_examples: 1810
download_size: 12131460
dataset_size: 43929245
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
license: mit
language:
- si
---
# Sinhala News Comment Sentiment Dataset

## Overview
This repository contains a dataset of Sinhala news comments categorized by sentiment (Positive, Negative, Neutral). To provide deeper context for advanced models, the dataset also includes metadata from the parent news article, including the article's title, author, publication date, and full body text.

**Note on Origin:** This Hugging Face repository was created purely for convenience. The original data was distributed as individual XML files. It has been parsed, cleaned, and formatted into standard `Dataset` objects here to allow seamless integration with the `datasets` library and modern LLM evaluation harnesses.

**All credit for the data collection, annotation, and original research belongs to the original authors.**

## Original Source & Credits
This dataset is the result of the research conducted by **S. Ranathunga** and **I. U. Liyanage**.

The original raw XML corpus and associated tagging scripts can be found in their official GitHub repository:
🔗 [theisuru/sentiment-tagger](https://github.com/theisuru/sentiment-tagger)

## Dataset Structure
The dataset has been restructured from XML into a flat, tabular format with the following features:
* `text` *(string)*: The text of the user's news comment.
* `label` *(ClassLabel)*: The sentiment of the comment.
* `0`: Negative
* `1`: Neutral
* `2`: Positive
* `article_body` *(string)*: The full text of the parent news article.
* `article_title` *(string)*: The title of the news article.
* `article_author` *(string)*: The author of the news article.
* `article_date` *(string)*: The publication date of the news article.

## Citation Information
If you use this dataset, code, or trained models derived from it, you are required by the original authors to cite their work.

**APA:**
> Ranathunga, S., & Liyanage, I. U. (2021). Sentiment Analysis of Sinhala News Comments. *Transactions on Asian and Low-Resource Language Information Processing, 20*(4), 1-23.

**BibTeX:**
```bibtex
@article{ranathunga2021sentiment,
title={Sentiment Analysis of Sinhala News Comments},
author={Ranathunga, Surangika and Liyanage, Isuru U},
journal={Transactions on Asian and Low-Resource Language Information Processing},
volume={20},
number={4},
pages={1--23},
year={2021},
publisher={ACM New York, NY}
}