VoiceAssistant-400K

该数据集包含了约47万个由GPT-4o专门生成的条目，提供了查询音频和回应的转录文本。在整理过程中，创建了一个清洁版本，去除了回应过长的实例。因此，规模上从47万个条目调整为43万组查询与回应的配对。该数据集的任务是语音生成及识别。

This dataset contains approximately 470,000 entries specially generated by GPT-4o, which provide transcribed texts of query audios and their corresponding responses. During the curation process, a cleaned version was created by filtering out instances with overly long responses. As a result, the dataset size was adjusted from 470,000 entries to 430,000 query-response pairs. The tasks targeted by this dataset are speech generation and speech recognition.