AudioSkills_af_next_events
收藏资源简介:
该数据集包含音频及其对应的元数据,共417个训练样本。每个样本包含16kHz采样率的音频、语言标识、来源、原始JSON结构、文本转录、标注文本、事件数量、唯一事件数量、唯一事件列表,以及一系列与Agent下一步交互相关的字段(包括提示、原始输出、推理轨迹、最终答案、状态、错误、时间戳)。该数据集可能适用于语音识别、事件检测、基于Agent的对话或推理任务。
This dataset contains audio and corresponding metadata, with a total of 417 training samples. Each sample includes audio at 16kHz sampling rate, language identifier, source, original JSON structure, text transcription, annotated text, number of events, number of unique events, list of unique events, and a series of fields related to the next interaction of the Agent (including prompts, raw outputs, reasoning traces, final answers, status, errors, timestamps). The dataset may be applicable to speech recognition, event detection, agent-based dialogue, or reasoning tasks.




