mHossain/Bengali_ParaPhrase_Corpus

BanglaParaCorpus是一个大规模的孟加拉语句子改写数据集，包含200万个句子对。该数据集旨在支持孟加拉语的句子改写生成、识别和评估任务。数据集采用多语言枢轴策略（孟加拉语→英语→孟加拉语）构建，并经过全面的过滤管道以确保语义一致性、流畅性和多样性。它包括正式（如新闻文章）和非正式（如社交媒体）文本，适用于广泛的下游NLP应用。

BanglaParaCorpus is a large-scale Bengali paraphrase generation dataset containing 2 million sentence pairs. It is designed to support paraphrase generation, recognition, and evaluation tasks in Bengali, a low-resource language. The dataset is built using a multilingual pivoting strategy (Bengali to English to Bengali) and a comprehensive filtering pipeline to ensure semantic consistency, fluency, and diversity. It includes both formal (e.g., news articles) and informal (e.g., social media) texts, making it suitable for a wide range of downstream NLP applications.