Trelis/fleurs-en-test

FLEURS英语测试集（Trelis）是一个用于自动语音识别（ASR）基准测试的私有评估集，包含647个话语样本。它源自Google的FLEURS数据集（en_us配置的test分割），FLEURS是一个覆盖102种语言的标准跨语言ASR基准，旨在评估语音通用表示的少样本学习性能。该测试集未经过滤或子采样，直接使用原始测试分割，包含音频字节、参考文本（转录）、原始FLEURS话语ID、说话者性别和持续时间（以秒计）等列。音频以原始字节形式存储以避免编解码器依赖，适用于ASR模型评估和比较。

FLEURS English Test Set (Trelis) is a private evaluation dataset for automatic speech recognition (ASR) benchmarking, containing 647 speech utterance samples. It is derived from Google's FLEURS dataset, specifically the test split of the en_us configuration. As a standard cross-lingual ASR benchmark covering 102 languages, FLEURS is designed to evaluate the few-shot learning performance of universal speech representations. This test set is unfiltered and unsampled, directly utilizing the original test split, and includes columns such as raw audio bytes, reference text (transcription), original FLEURS utterance IDs, speaker gender, and duration in seconds. The audio is stored as raw bytes to avoid codec dependencies, making it suitable for ASR model evaluation and comparison.