LaboroTVSpeech

LaboroTVSpeech是由日本东京的Laboro.AI, Inc.和东京大学合作创建的大型日语语音数据集，包含超过2000小时的语音数据，来源于日本电视台的录音及其字幕。数据集的创建过程自动化，通过迭代工作流程从电视录音中提取匹配的音频和字幕片段。该数据集主要用于训练自动语音识别（ASR）系统，特别是在处理多样化的语音和背景噪声方面表现出色。随着时间的推移，数据集的规模将持续增加，为ASR系统的研究提供丰富的资源。

LaboroTVSpeech is a large-scale Japanese speech dataset co-developed by Laboro.AI, Inc. based in Tokyo, Japan and The University of Tokyo. It contains over 2,000 hours of speech data sourced from Japanese TV station recordings and their matching subtitles. The dataset is built through an automated iterative workflow that extracts aligned audio and subtitle segments from television recordings. Primarily designed for training automatic speech recognition (ASR) systems, this dataset performs excellently in handling diverse speech varieties and background noise. Over time, the scale of the dataset will continue to expand, providing rich resources for ASR system research.