YTCommentQA
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/lgresearch/YTCommentQA
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为YTCommentQA，其中包含了从YouTube自然生成的提问，这些问题根据回答的可能性以及回答所需的模态（视觉、文本或两者兼备）进行了分类。数据集还包括了回答可能性的类别：无法回答、仅通过视觉可回答、仅通过文本可回答，以及当提供视觉和文本时才能回答。该数据集包含了来自2004个视频的2332个问题，其任务是进行回答可能性分类。
The dataset, named YTCommentQA, consists of naturally generated questions sourced from YouTube. These questions are categorized based on two dimensions: answerability and the modality required for answering (visual, textual, or both). The answerability categories included in the dataset are unanswerable, answerable only via visual information, answerable only via textual information, and answerable only when both visual and textual information are provided. The dataset comprises 2,332 questions from 2,004 videos, and its associated task is answerability classification.
