fatymahaly/Urdu_Sentement

Roman Urdu Sentiment Analysis Dataset

A clean, structured dataset for sentiment analysis in Roman Urdu, split into training and testing sets. This dataset is optimized for quick integration with the Hugging Face datasets library and is ideal for fine-tuning text classification models (such as DistilBERT, mBERT, or XLM-RoBERTa) to handle Roman Urdu customer feedback, reviews, and social media text.

Dataset Structure

The dataset contains short user reviews and text samples labeled across three sentiment classes: positive, negative, and neutral.

Data Fields

• text: The input sentence or review written in Roman Urdu.
• label: The sentiment category (positive, negative, or neutral).

Data Splits

• Training Set (train.csv): 100+ samples
• Test Set (test.csv): 100+ samples

Quick Start (Python)

You can load this dataset directly into your environment using the Hugging Face datasets library:

from datasets import load_dataset

# Load local train and test csv files
dataset = load_dataset(
    "csv", data_files={"train": "train.csv", "test": "test.csv"}
)

print(dataset)
print(dataset["train"][0])