KoelLabs/real-world-noise-through-zoom

RWNTZ（Real-World Noise Through Zoom）数据集是一个包含真实世界噪音环境下的自动语音识别的数据集。该数据集包含5种不同的真实噪音环境（卧室、拥挤的房间、背景音乐、雨声、有车辆的路面），2位不同的说话人，以及不同麦克风距离（6英寸和24英寸）的录音。共有32个样本，每个样本包含不同的短语，通过Zoom录制以模拟现实世界的语言田野调查场景。每个样本都有手动验证的单词级别转录和基于g2p的音素转录，以及使用多种基于Wav2Vec2模型的音频到音素的转录。

The RWNTZ (Real-World Noise Through Zoom) dataset is an automatic speech recognition dataset that includes real-world noise conditions. The dataset consists of 5 different real noise environments (bedroom, crowded room, background music, rain, road with cars), 2 different speakers, and recordings at various microphone distances (6 inches and 24 inches). There are a total of 32 samples, each containing different phrases, recorded through Zoom to simulate real-world linguistic fieldwork scenarios. Each sample has manually verified word-level transcriptions and g2p phoneme transcriptions, as well as audio-to-phoneme transcriptions using a variety of Wav2Vec2-based models.