Banglish

孟加拉语和英语语音命令的双语数据集 由于殖民影响，孟加拉语口语采用了许多英语单词。在会话孟加拉语中，混合使用英语和孟加拉语是很常见的。这种在会话语言中普遍存在的现象被称为语码转换（CS）。 CS被定义为在一次对话中两种语言之间的连续交替。因此，在孟加拉语自然语言处理中，通常需要将单个基本命令映射到它的许多不同变体——用英语和孟加拉语的多种混合语言。为了促进这一点，我们策划了一个以常见浏览器命令为中心的数据集。

Bilingual Dataset of Bengali and English Voice Commands
Due to colonial influence, spoken Bengali has adopted numerous English words. Mixing English and Bengali in conversational Bengali is highly common. This prevalent phenomenon in conversational language is termed code-switching (CS), which is defined as the continuous alternation between two languages within a single conversation. Consequently, in Bengali natural language processing (NLP), it is typically necessary to map a single basic command to its various distinct variants across diverse mixed combinations of English and Bengali. To facilitate this requirement, we curated a dataset centered on common browser voice commands.

搜集汇总

背景与挑战

背景概述

该数据集是面向孟加拉语和英语混合语音命令的双语资源，专门针对语码转换场景设计，旨在支持浏览器指令的多种语言变体识别。由Google AI Research于2021年发布，数据量为893.0B。