513小时日语语音数据_对话（电话）【数据堂】

513小时日语自然对话电话信道语音数据由800余名发音人参与录制，以自然方式进行交流，针对给定的数个话题自由发挥，领域广泛，语音自然流利，符合实际对话场景。人工转写文本，准确率高

513-hour Japanese natural conversational telephone-channel speech dataset recorded by over 800 speakers. The conversations were conducted naturally, with participants speaking freely on several given topics spanning a wide range of domains. The speech is natural and fluent, conforming to real-world conversational scenarios. The dataset is paired with manually transcribed texts of high accuracy.

搜集汇总

背景与挑战

背景概述

该数据集包含513小时的日语自然对话电话录音，由878名日本人参与录制，采用人工转写文本，句准确率达95%。数据格式为8kHz单声道WAV文件，适用于语音识别和声纹识别等应用场景。