Golos

本文介绍了一种新颖的俄语语音数据，称为Golos，这是一种适合语音研究的大型语料库。该数据集主要由在众包平台上进行人为注释的记录音频文件组成。音频的总持续时间约为1240小时。我们可以免费下载语料库，以及在此计算机上准备的带有CTC损失的声学模型。

This paper introduces a novel Russian speech dataset named Golos, a large-scale corpus suitable for speech research. This dataset mainly consists of recorded audio files manually annotated on crowdsourcing platforms. The total duration of all audio recordings is approximately 1,240 hours. The corpus, along with the acoustic models trained with CTC loss prepared on this computer, is available for free download.

搜集汇总

背景与挑战

背景概述

Golos是一个大型俄语语音数据集，包含约1240小时的众包注释音频文件，适用于语音研究，可免费获取并附带预训练的声学模型。该数据集由俄罗斯联邦储蓄银行于2021年发布。