SURE (Multimodal Recommendation Dialog with SUbjective PREference)
收藏资源简介:
SURE数据集是由北京邮电大学创建的一个大规模多模态对话数据集,包含12,180个购物对话,旨在研究基于用户主观偏好的商品推荐。数据集通过两个阶段的构建过程,确保了对话质量和语言多样性。SURE数据集详细标注了用户的主观偏好和推荐行为,提供了丰富的资源用于评估主观偏好理解和对话策略。该数据集主要用于构建能够理解用户主观偏好并提供有效推荐的对话代理,解决在实际购物对话中用户偏好表达多样性的挑战。
The SURE dataset is a large-scale multimodal dialogue dataset created by Beijing University of Posts and Telecommunications, which contains 12,180 shopping dialogues and is aimed at investigating product recommendation based on users' subjective preferences. It is constructed through a two-stage workflow to ensure dialogue quality and linguistic diversity. The SURE dataset is meticulously annotated with users' subjective preferences and recommendation behaviors, serving as a rich resource for evaluating subjective preference understanding and dialogue strategies. This dataset is primarily utilized to develop dialogue agents that can comprehend users' subjective preferences and provide effective recommendations, tackling the challenge of diverse expression of user preferences in real-world shopping conversations.

- 1Multimodal Recommendation Dialog with Subjective Preference: A New Challenge and Benchmark北京邮电大学 · 2023年



