Vikhrmodels/RuASRBenchmark

RuASRBenchmark是一个用于评估俄语自动语音识别（ASR）系统质量的基准数据集。该数据集整合了多个开源的俄语语音数据集，提供了统一的测试格式。数据集包括俄语音频书籍及其转录、Mozilla Common Voice的俄语众包录音、清晰风格的演讲录音、录音室录制的书籍、合成的俄语语音数据以及非专业麦克风录制的用户语音。

RuASRBenchmark is a benchmark dataset for evaluating the quality of Russian Automatic Speech Recognition (ASR) systems. The dataset combines several open-source Russian speech datasets and provides a unified format for testing. It includes Russian audiobooks and their transcriptions, crowdsourced Russian recordings from Mozilla Common Voice, clear-style presentation recordings, studio-recorded books, synthetic Russian speech, and user speech recorded on non-professional microphones.