CrowdSpeech and Vox DIY: Benchmark Dataset for Crowdsourced Audio Transcription

We collect and release CrowdSpeech — the first publicly available large-scale dataset of crowdsourced audio transcriptions. e show its applicability on an under-resourced language by constructing VoxDIY — a counterpart of CrowdSpeech for the Russian language.

本研究收集并发布了CrowdSpeech——首个公开可用的大规模众包音频转录数据集。本研究通过构建针对俄语的CrowdSpeech对应数据集VoxDIY，验证了该数据集在低资源语言场景下的应用可行性。