数据堂—500小时闽南语自然对话手机采集语音数据

500小时闽南语自然对话手机采集语音数据由约1000名本土闽南语使用者参与录制，录音人来自泉州、漳州、厦门地区，男女比例均衡，覆盖多个年龄段，500小时闽南语自然对话手机采集语音数据无预设语料，为确保对话的流畅自然，由录音人根据自己熟悉的话题展开对话并录制。

This 500-hour mobile-collected speech dataset features natural Minnan conversations, recorded by approximately 1,000 native Minnan speakers from Quanzhou, Zhangzhou and Xiamen. The participant cohort has a balanced gender ratio and covers multiple age groups. No pre-designed corpus was utilized for this dataset: to ensure fluent and natural dialogues, speakers were instructed to initiate conversations based on topics they are familiar with.

搜集汇总

背景与挑战

背景概述

该数据集包含500小时通过手机采集的闽南语自然对话语音，用于测试语音识别模型。数据由约1000名来自泉州、漳州、厦门的母语者录制，基于熟悉话题进行无脚本对话，格式为16kHz的.wav音频。