PhotoBook

该数据集是一个大规模的、以视觉为基础、面向任务的英语对话集合，旨在研究对话过程中共享对话历史是如何逐渐累积的。它包含了2,506个完成的游戏，总计164,615条发言和130,322个动作。参与者必须是英语母语者，并且需要通过仅包含三张图片的热身轮次。该数据集使得研究者能够分析在多轮沟通任务中任务效率和语言使用情况。其规模涵盖了2,506个游戏、164,615条发言、130,322个动作以及11,805个独特词汇。该数据集的任务是视觉基础对话。

This dataset is a large-scale, vision-grounded, task-oriented English dialogue collection designed to investigate the gradual accumulation of shared conversation history during conversational interactions. All participants must be native English speakers and complete a warm-up round involving only three images. This dataset enables researchers to analyze task efficiency and language usage patterns within multi-turn communication tasks. In total, it encompasses 2,506 completed games, 164,615 utterances, 130,322 actions, and 11,805 unique lexical items. The core task of this dataset is vision-grounded dialogue.