susameddin/Sympatheia-18k
收藏资源简介:
Sympatheia-18k是一个情感感知的语音对话数据集,专为共情语音合成研究设计。它包含18,000个查询-响应对,覆盖12种情感类别(包括愤怒、焦虑、满足、厌恶、兴奋、沮丧、快乐、中性、放松、悲伤、惊讶和疲劳)。每个对都配有合成音频和文本转录。数据集分为两个子集:Emotional子集包含情感查询和情感匹配的响应,而Neutral子集包含中性查询,每个查询与12种情感目标响应配对。数据集还包括音频文件(WAV格式)、元数据文件(JSONL格式,包含查询和响应文本对以及情感标签)和预编码的音频令牌(用于Sympatheia模型)。情感类别映射到效价-唤醒值,以支持情感感知的语音合成任务。
Sympatheia-18k is an emotion-aware spoken dialogue dataset for empathetic speech synthesis research. It contains 18,000 query–response pairs across 12 emotion categories (Angry, Anxious, Content, Disgusted, Excited, Frustrated, Happy, Neutral, Relaxed, Sad, Surprised, Tired), each accompanied by synthesized audio and text transcripts. The dataset is structured into two subsets: the Emotional subset includes emotional queries with emotionally-matched responses, and the Neutral subset consists of neutral queries each paired with 12 emotion-targeted responses. It provides audio files (in WAV format), metadata files (in JSONL format, containing text pairs and emotion labels), and pre-encoded audio tokens (for use with the Sympatheia model). Emotion categories are mapped to valence–arousal values to facilitate emotion-aware speech synthesis tasks.



