Transliteration Sentence Dataset

A transliteration sentence is like writing the same words but using different letters that sound the same. It helps people who speak different languages understand each other better. This dataset, drawn from 12 varied datasets initially intended for tasks such as sentiment analysis, hate speech detection, social media analysis, and review classification, endeavors to encompass a wide array of linguistic subtleties and fluctuations inherent in real-world language usage. Each data instance was meticulously labeled based on the language of the sentences. From this amalgamation of datasets, we curated a dataset comprising 65,473 instances, comprising 19,859 Bangla, 17,309 Hindi, 17,000 English, and 11,305 Tamil data instances, specifically tailored for transliteration sentence identification.

音译句（transliteration sentence）是指采用发音相同的不同字母来书写同一词汇的文本形式，可帮助不同语言使用者实现更高效的跨语言沟通。本数据集源自12个初始用于情感分析、仇恨言论检测、社交媒体分析及评论分类等任务的多样化数据集，旨在涵盖真实语言使用场景中固有存在的各类语言细微差异与动态变化。每一条数据样本均依据句子所属语言进行了精细标注。我们从这些整合后的数据集当中筛选构建出本数据集，共计包含65473条数据样本，其中孟加拉语（Bangla）样本19859条、印地语（Hindi）样本17309条、英语样本17000条以及泰米尔语（Tamil）样本11305条，专为音译句识别任务量身打造。