minato-ryan/emilia-yodas-cjk-xcodec2-example

该数据集包含四个配置：ja_long、ja_short、ja_skipped和ja_very_long。每个配置都包含以下特征：用户ID（uid）、dnsmos评分、时长（duration）、语言（language）、电话数量（phone_count）、说话者（speaker）、文本（text）和序列ID（ids）。ja_skipped配置还包括音频信息，如音频数组（array）、音频路径（path）和采样率（sampling_rate）。数据集被划分为训练集（train），每个配置的训练集大小不同。The dataset consists of four configurations: ja_long, ja_short, ja_skipped, and ja_very_long. Each configuration includes features such as user ID (uid), dnsmos score, duration, language, phone count (phone_count), speaker, text, and sequence ID (ids). The ja_skipped configuration also includes audio information such as audio array (array), audio path (path), and sampling rate (sampling_rate). The dataset is split into training sets (train), and the size of the training set varies for each configuration.