ggfox00000/stt-mediaspeech-test

MediaSpeech法语测试集是用于法语自动语音识别（ASR）基准测试的数据集，包含来自广播、电视和播客的短片段。数据集包含2498个话语片段，每个片段约10秒，总时长为10小时。音频格式为FLAC 16 kHz mono PCM_16，语言为法语，许可证为CC-BY-4.0。数据集的结构包括id（UUID标识符）、transcript（法语参考转录文本）、duration_sec（片段时长，单位为秒）和audio（包含路径、数组和采样率的音频数据）。数据集来源于MediaSpeech（MTS AI），是OpenSLR 108的一部分，包含四种语言（阿拉伯语、法语、西班牙语和土耳其语），本数据集仅包含法语部分。

The MediaSpeech French test split is a dataset for benchmarking French automatic speech recognition (ASR), containing short segments from broadcasts, TV, and podcasts. The dataset includes 2498 utterances, each approximately 10 seconds long, totaling 10 hours of audio. The audio format is FLAC 16 kHz mono PCM_16, and the language is French. The license is CC-BY-4.0. The dataset structure includes fields for id (UUID identifier), transcript (French reference transcription), duration_sec (segment duration in seconds), and audio (audio data containing path, array, and sampling rate). The dataset originates from MediaSpeech (MTS AI) and is part of OpenSLR 108, which includes four languages (Arabic, French, Spanish, and Turkish), with this dataset containing only the French portion.