Bengali Voice Commerce Audio Dataset for Automatic Speech Recognition

This meticulously curated dataset presents a comprehensive collection of Bengali speech recordings meticulously compiled for the advancement of voice-enabled commerce applications. Consisting of 3344 high-quality audio samples, this dataset encapsulates diverse voice recordings articulating product names and quantities commonly encountered in retail and e-commerce settings. Each recording, encompassing a spectrum of environmental conditions ranging from bustling locales to controlled studio settings, offers a rich repository for developing and enhancing Automatic Speech Recognition (ASR) systems. Alongside the audio files, meticulously annotated Bengali and English transcripts stored in a CSV format accompany each recording, ensuring comprehensive linguistic documentation. Featuring contributions from 80 native Bengali speakers, including a balanced representation of 45 male and 35 female speakers, this dataset facilitates robust research in Bengali ASR, fostering innovation in voice commerce technology and inclusive digital interactions. Researchers and practitioners in machine learning, natural language processing, and human-computer interaction can leverage this dataset to propel advancements in voice-driven applications and accessibility initiatives.

本经精心遴选编撰的数据集收录了全面的孟加拉语语音录音样本，专为推动语音电商应用的发展而汇编。该数据集共包含3344条高质量音频样本，涵盖零售与电商场景中常见的商品名称及数量表述的多样化语音录音。每条录音均涵盖从喧闹公共场所到受控演播室等多元环境条件，可为开发与优化自动语音识别（Automatic Speech Recognition, ASR）系统提供丰富的资源储备。配套音频文件之外，每条录音均附带以CSV格式存储的经精细标注的孟加拉语与英语转写文本，确保提供完整的语言学文档资料。该数据集的语音数据来自80名孟加拉语母语者，其中男性45名、女性35名，性别分布均衡，可为孟加拉语ASR研究提供坚实支撑，助力语音电商技术创新与包容性数字交互的发展。机器学习、自然语言处理及人机交互领域的研究人员与从业者，可借助该数据集推动语音驱动应用及无障碍数字化倡议的研发进展。