Minuri/

sinhala-perplexity-test-dataset

	Sinhala Perplexity Test Dataset




	
	Dataset Description

This dataset contains 500 sentence pairs in Sinhala, provided in two scripts: Unicode Sinhala and Romanized Sinhala. It is intended for evaluating perplexity and benchmarking language models on Sinhala text. This dataset was introduced and used in the following benchmark study:

Rajapakse, M., & Weerasinghe, R. (2025). Script Sensitivity: Benchmarking Language Models on Unicode, Romanized and Mixed-Script Sinhala.… See the full description on the dataset page: https://huggingface.co/datasets/Minuri/sinhala-perplexity-test-dataset.