Bangla News Dataset

A corpus on Bangla newspaper articles created using a custom web crawler containing 12 different topics. The total number of word tokens in this dataset is 28.5+ million. The number of unique words is around 3% of the entire vocabulary of the dataset. The Dataset is imbalanced. 20% of the dataset was separated as a held-out dataset.

本数据集为基于定制网络爬虫构建的孟加拉语报纸文章语料库，涵盖12个不同主题。该数据集的词元（Token）总数量达2850余万。数据集中的独特词汇数量约占数据集总词汇量的3%。该数据集存在类别不平衡问题。已将数据集的20%划分为留出数据集（held-out dataset）。