LibriSQA
收藏资源简介:
LibriSQA是一个专为大型语言模型优化的新型口语问答数据集,由上海交通大学协同媒体创新中心创建。该数据集包含214,000个口语问答对,覆盖广泛的主题,分为两部分:第一部分设计用于自然对话格式,第二部分专注于多项选择题及其分析段落。LibriSQA旨在通过提供自由形式的开放式问答,推动大型语言模型在多模态任务中的理解和交互能力。该数据集的应用领域包括提升语言模型的口语理解和生成能力,以及在自动语音识别任务中的应用。
LibriSQA is a novel spoken question answering dataset optimized for large language models, created by the Collaborative Media Innovation Center of Shanghai Jiao Tong University. This dataset contains 214,000 spoken question-answer pairs covering a wide range of topics, and is divided into two parts: the first part is designed for natural conversation formats, while the second part focuses on multiple-choice questions and their analytical passages. LibriSQA aims to advance the understanding and interaction capabilities of large language models in multimodal tasks by providing free-form open-ended question answering. Its application areas include improving the spoken language understanding and generation capabilities of language models, as well as applications in automatic speech recognition tasks.




